ArticlePDF Available

Smart Cupboard for Assessing Memory in Home Environment


Abstract and Figures

Sensor systems for the Internet of Things (IoT) make it possible to continuously monitor people, gathering information without any extra effort from them. Thus, the IoT can be very helpful in the context of early disease detection, which can improve peoples’ quality of life by applying the right treatment and measures at an early stage. This paper presents a new use of IoT sensor systems—we present a novel three-door smart cupboard that can measure the memory of a user, aiming at detecting potential memory losses. The smart cupboard has three sensors connected to a Raspberry Pi, whose aim is to detect which doors are opened. Inside of the Raspberry Pi, a Python script detects the openings of the doors, and classifies the events between attempts of finding something without success and the events of actually finding it, in order to measure the user’s memory concerning the objects’ locations (among the three compartments of the smart cupboard). The smart cupboard was assessed with 23 different users in a controlled environment. This smart cupboard was powered by an external battery. The memory assessments of the smart cupboard were compared with a validated test of memory assessment about face–name associations and a self-reported test about self-perceived memory. We found a significant correlation between the smart cupboard results and both memory measurement methods. Thus, we conclude that the proposed novel smart cupboard successfully measured memory.
Content may be subject to copyright.
Smart Cupboard for Assessing Memory in
Home Environment
Franks González-Landero 1, Iván García-Magariño 2,* , Rebecca Amariglio 3,4 and
Raquel Lacuesta 5,6
1Edison Desarollos, 44002 Teruel, Spain;
2Department of Software Engineering and Artificial Intelligence, Complutense University of Madrid,
28040 Madrid, Spain
3Harvard Medical School, Harvard University, Boston, MA 02115, USA;
4Massachusetts General Hospital, Boston, MA 02114, USA
Department of Computer Science and Engineering of Systems, University of Zaragoza, 44003 Teruel, Spain;
6Instituto de Investigación Sanitaria Aragón, University of Zaragoza, 50009 Zaragoza, Spain
*Correspondence:; Tel.: +34-913-947-643
Received: 26 April 2019; Accepted: 31 May 2019; Published: 4 June 2019
Sensor systems for the Internet of Things (IoT) make it possible to continuously monitor
people, gathering information without any extra effort from them. Thus, the IoT can be very helpful in
the context of early disease detection, which can improve peoples’ quality of life by applying the right
treatment and measures at an early stage. This paper presents a new use of IoT sensor systems—we
present a novel three-door smart cupboard that can measure the memory of a user, aiming at detecting
potential memory losses. The smart cupboard has three sensors connected to a Raspberry Pi, whose
aim is to detect which doors are opened. Inside of the Raspberry Pi, a Python script detects the
openings of the doors, and classifies the events between attempts of finding something without
success and the events of actually finding it, in order to measure the user’s memory concerning the
objects’ locations (among the three compartments of the smart cupboard). The smart cupboard was
assessed with 23 different users in a controlled environment. This smart cupboard was powered
by an external battery. The memory assessments of the smart cupboard were compared with a
validated test of memory assessment about face–name associations and a self-reported test about
self-perceived memory. We found a significant correlation between the smart cupboard results and
both memory measurement methods. Thus, we conclude that the proposed novel smart cupboard
successfully measured memory.
Keywords: IoT; memory loss; e-healthcare; Alzheimer’s; door sensors
1. Introduction
Dementia is the progressive loss of cognitive functions due to brain damage or disorder.
Among dementia types, one of the most well-known and widespread diseases is Alzheimer’s disease,
and the number of people who will suffer from this disease is estimated to reach 131.5 million by
2050 [
]. This disease hampers daily life activities such as recognizing faces and remembering names,
places, and positions [
]. Potentially 131.5 million people could put themselves at risk if they start
developing these symptoms without the proper cautionary measures and palliative treatments.
There is no cure for Alzheimer’s disease, but there are palliative treatments. Some of these are
medicines, but none of them has been proven to stop the progression of this disease [
]. Other examples
are psychosocial interventions, and these involve stimulation-oriented treatments with art, music,
Sensors 2019,19, 2552; doi:10.3390/s19112552
Sensors 2019,19, 2552 2 of 21
animals, or other recreational activities; however, the efficacy of these treatments remains uncertain [
The last of the palliative treatments is caregiving, which is probably the safest one, but it has a great
impact in health economics for the necessary resources—mainly caregivers, but also measurement of
the state of the disease for suitable caregiving, tagging items in houses, and the maintenance of feeding
tubes in the case of eating problems.
On the other hand, the literature claims that Alzheimer’s disease represents the main cause of
neurodegenerative dementia in the population aged over 60 years old, with an estimated prevalence
of 5–7% [
]. People increase the probability of starting to suffer from Alzheimer’s disease when
getting older without noticing. If anyone suffers from Alzheimer’s disease, they need to receive some
treatment—the sooner the better.
At present, most people live surrounded by technology, including Internet of Things (IoT) sensor
objects that can collect, pre-process, and analyze continuous streams of data (e.g., weather, traffic,
finance, and health data). One of the most common goals of the IoT is to enhance the quality of life.
IoT technology can track the interactions between quotidian objects and persons, or even among
objects, contributing to the digitalization of the physical world. Another goal of the IoT is to connect
and synchronize traditional utensils through the Internet in order to deliver a service more efficiently.
In this way, all elements that used to connect in a close circuit are now connected through a network,
increasing their utility [
]. IoT sensor systems now allow the connection of physical objects so that
remote services can be provided through the Internet by analyzing the data from these sensors.
Thus, once the hardware of the IoT sensors is installed, programmers can provide new functionalities
by developing new software based on different analyses.
In this paper we present a novel sensors system based on the IoT aimed at detecting memory
losses for the early detection of some neurodegenerative diseases, by continuously assessing the
memory of the user and notifying them when appropriate. Among other features of the system, we can
highlight that this measurement method does not require any additional effort from the user, and is
continuous. Users only need to go about their daily routine. The sensors system was installed in a
cupboard, converting it into an IoT smart cupboard (SC). We used a normal cupboard, such as one that
most readers could find at their home or in their kitchen. The SC had some door sensors connected to
a Raspberry Pi, programmed to analyze the signals and measure the memory.
The main contribution of the current work over the related works of other authors is its
presentation of a low-cost solution that can monitor users in their daily lives for measuring memory
with the potential of detecting diseases with memory impairments, without needing qualified staff.
In addition, the mechanism is novel, as this is the first work that presents a novel SC for this purpose,
and is based on magnetic door sensors with very low prices. This work extends our previous work
about IoT collaboration exemplified with a SC prototype [
]. The contribution of the current work over
the previous one lies in the use of a more advanced three-door SC prototype that measures memory.
This wa proved with experimental results obtained from 23 participants in which the SC measurements
correlated with a validated memory test and another test about self-perceived memory.
The current work is organized as follows. The next section reviews the existing related works
considering the common technologies in this field. Section 3describes the design of the proposed SC
to assess memory and all its features. Section 4describes the conducted experiments and the user tests
for validating the system. Section 5presents the main results. Finally, Section 6discusses the results,
draws conclusions, and depicts some future lines of research.
2. Related Work
The research community is actively involved in the topic of this work due to the consequences
of memory losses on the wellbeing of patients and their social environments. The goal is to reduce
their economic impact on the society due to treatment costs. IoT technology allows the interconnection
of small low-cost devices practically anywhere. These devices can monitor health indicators and the
Sensors 2019,19, 2552 3 of 21
behaviors of people. There are many works and projects on this topic, and this section introduces the
most relevant ones.
Some projects present solutions involving Alzheimer’s disease and Raspberry Pi. For instance,
Nonavinakere et al. [
] developed a system that recognizes a person’s face and tells the user the name
of that person and the relationship they have with them. The tool was developed thinking about users
with Alzheimer’s disease. The system was tested using three different platforms; one of them was a
Raspberry Pi 3, and although it was not the fastest platform, it was the most accurate, since it was able
to detect a person within certain limits. Crema et al. [
] proposed an embedded platform-based system
for early detection of Alzheimer’s disease through transcranial magnetic stimulation (TMS). TMS is a
non-invasive way to stimulate the cerebral cortex in order to address Alzheimer’s disease. This system
was formed by a magnetic stimulus generator, an electric stimulus generator, a field-programmable
gate array, and a Raspberry Pi. The goal was to introduce an alternative technique that supported
the early detection of Alzheimer’s with reduced costs, and provided results that were suitable for
medical interpretation. Narendiran et al. [
] developed a cognitive assistance system for smart
homes. The main aims of the project were (a) to simulate the progression of Alzheimer’s-type
dementia by evaluating performance in the execution of an activity of daily living and (b) to provide
support for impaired people who need help in daily activities such as preparing a cup of coffee.
The system used a camera connected to Raspberry Pi. The camera provided live images to the
Raspberry, whose contents were the patient inside the home environment. While the Raspberry was
receiving images, it assessed the performance of a task and provided feedback to the patient about
their performance. For instance, when a patient forgot some step of some task, the system reminded
the patient about this step. Ishii et al. [
] designed an early-detection system for dementia using
the Machine-to-Machine (M2M)/IoT platform. The system was formed by sound sensors, motion
sensors, pressure sensors, an Arduino board per sensor, a Raspberry Pi board, an M2M server, and the
corresponding analysis software. The authors assessed several activities and behaviors of a person
inside their home. Several sensors were set up in the home environment, including outdoors near the
home and inside some rooms (e.g., bedrooms, bathrooms). Each Arduino board connected to a sensor
sent information to the Raspberry Pi. This had two functions; the first was to send information to the
M2M server, and the second was to analyze the collected information through the analysis software
in order to determine early symptoms of Alzheimer’s disease. It is worth highlighting the mixed
use of Arduino and Raspberry Pi boards. Kristalina et al. [
] kept in mind one of consequences of
memory impairment, which is that a certain person can forget where they are. In order to address
this handicap, they developed a system that involved a Raspberry Pi and an iBeacon. This was a
tracking system for patients with memory impairment. Each patient carried an iBeacon device that
was responsible for sending the ID and signal strength to the Raspberry Pi and then to the server in
order to convert the information to a distance. Due to the amount of noise, it was possible that the
obtained distance data did not match with current patient position, so the authors applied the Kalman
method in order to estimate the distance between devices. This system was assessed at the Dr. M.
Soewandhie hospital, and their tests showed an average percentage measurement error of 7.01% in the
actual patient position. Chavan and Chavan [
] proposed a novel system for fall detection in elderly
people. They benefited from the new features of Raspberry Pi 3 with respect to the previous version
(i.e., Wi-Fi connection) for the creation of a new system with wearables. The system was formed by a
laptop, a Raspberry Pi 3, accelerometers, a heart rate sensor, and a temperature sensor. The sensors
sent information to the Raspberry, and this determined if there had been a fall. If so, a text message
was sent to a mobile device. Paul et al. [
] described a low-cost system for monitoring patients, which
was formed by several sensors, a Raspberry Pi, a database, and an application. The system’s aim was
to collect patient data (e.g., electrocardiogram signal, blood pressure signal, heart rate signal, blood
oxygenation, temperature) in order to send them to the patient’s doctor. The system was set up in
Bangladesh with success.
Sensors 2019,19, 2552 4 of 21
Although they do not make use of a Raspberry Pi, the following projects addressed Alzheimer’s
disease with other IoT elements. Chong et al. [
] proposed a system in order to predict a potential
Alzheimer’s medical condition. They used a room with movement sensors inside it and analyzed
the data obtained from 20 elderly persons by means of five sensors over the course of six months.
The results provided three key factors in order to perform a prediction: excess activity levels, sleeping
patterns, and repetitive actions. These factors were useful for predicting the early warning signs
of Alzheimer’s, and allowed the authors to provide recommendations to caregivers based on the
prediction analyses. Navarro et al. [
] developed a fuzzy adaptive cognitive stimulation therapy
generation system for Alzheimer’s patients. The aim of the system was to reduce the cognitive
burden of care workers and therapists. The system assessed patient behavior through several activities
and even through their voice tone and their phrases. This system used the Mente Activa software,
whose aim was to provide computer-assisted cognitive therapy. The authors demonstrated the
enhancement of patients with the experiments with their system. Finally, Roopaei and Jane [
] focused
on another aspect related to memory loss, which was the ability to recognize familiar faces. The authors
developed a platform to support patients who suffered from face perception impairment with an
assistive intelligence device. The system included an algorithm that recognized a face among entries
in a face dataset. The algorithm used deep learning to recognize patterns in faces and match them.
Regarding IoT, the authors proposed to use glasses in order to let the user know who was in front of
them as well as their relationship.
Table 1depicts the main differences and similarities of the current work with the most related
ones. As one can observe, the current work is the only one that has all the following four features at
the same time: (a) it does not need qualified staff, and hence anybody can use it without previous
experience; (b) it has the potential to measure memory by just analyzing the daily activities of users;
(c) it is a low-cost solution for monitoring; and (d) it can detect symptoms of memory disease in early
stages. The most similar work is the one by Ishii et al. [
], as it also has the potential to measure
memory and conduct the early detection of memory-impairment related diseases by analyzing daily
activities without requiring qualified staff. Even though, the current work has all these features, it is
also low cost, thanks to the novel mechanism based on a SC with very low-cost magnetic door sensors.
Considering all the related works presented in this section, we also noticed a gap in the literature
about using pieces of IoT-enabled furniture for monitoring the memory of users for the early detection
of memory-impairment diseases. The current approach covers this gap in the literature by presenting
an IoT SC, built with low-cost magnetic door sensors, introduced in the next section.
Sensors 2019,19, 2552 5 of 21
Table 1. Comparison between the current work and the most closely related ones.
Question Current
et al. [8]
et al.
et al. [10]
et al.
et al.
Does this work use IoT? 3- - - 3
Does this work use
Raspberry Pi?
3 3 3 3 3 3
Does this work present a
low-cost solution for health
3 3 - - - 3
Does this system use
wearable devices?
-3 3 - - 3
Can this solution be applied
without qualified staff?
3- - - 3
Does this solution measure
3- - - 3
Does this solution measure
cardiac measures? (heart rate,
heart rate variability)
- - - - - 3
Does this solution measure
- - - - - 3
Does this solution have
the potential to measure
any health indicator by just
analyzing the daily activities
of users?
3- - 3 3 3
Does this solution have the
potential to measure memory
by just analyzing the daily
activities of users?
3- - 3 3
Could this solution help to
detect memory-impairment
diseases at an early stage?
3-3 3 3
3. Smart Cupboard for Assessing Memory
In Figure 1, the reader can observe a picture that depicts the overall experiment described in
this paper. This paper also presents the design of the SC, the assessment method of the SC, and the
experimentation with users. The core of the system is formed by a Raspberry Pi model 3 B+, with CPU
1.4 GHz 64-bit quad-core ARMv8, 1 GB Memory (SDRAM) (shared with GPU), 17x GPIO and HAT ID
bus, 5 V through MicroUSB or a GPIO header. A lithium-ion battery accompanies the Raspberry Pi,
which facilitated the setting up of the system inside the cupboard for the experiments, since a wire
connected to power was not necessary. This provided the possibility of installing the sensors system in
a cupboard without needing a nearby socket. The autonomy of the battery was 9 h, and we considered
that this was enough to assess our system in controlled environments with users. The SC also had
magnetic door sensors. The cupboard selection was made based on certain features. The requirements
that the furniture must have according to our controlled experiments were (a) to be placed in a
kitchen, (b) to have three compartments of the same size (to avoid memory techniques based on the
size of objects and compartments), and (c) that each compartment could hold 5 to 10 items without
overlapping (to facilitate the acquisition phase based on observation). We used an Excellway MC-38
wired magnetic alarm system door window sensor switch with screw provided by Banggood. Figure 2
shows this magnetic door sensor. Each sensor was composed of two parts; one was attached to the
cupboard structure and connected to the Raspberry Pi, and the other was attached to the door, such
that both parts were together when the door was closed and apart when it was open. Each sensor
closed the circuit when both parts were together (or very near to each other). For our system, we used
three pairs of sensors and these were set up in the three doors of the cupboard. A protoboard and
Sensors 2019,19, 2552 6 of 21
jumper wires were used to connect the Raspberry Pi and door sensors, and the schematic design is
presented in Section 3.1. We wrote a Python script in order to manage the proposed SC. The script
assessed the memory of users based on the analysis of the signals of door sensors, with the algorithm
described in Section 3.2.
Figure 1. Overview of the smart cupboard and the experiments.
Figure 2. Door sensor.
3.1. Schematic Design of the Smart Cupboard
The door sensors and the Raspberry Pi were connected through jumper wires and a protoboard,
and this section explains in detail how these elements were connected. The Raspberry Pi had a pin
series placed on a side of the board, called GPIOs (general-purpose inputs/outputs). They performed
multiple input/output operations for different purposes. The Raspberry Pi model used in this work
had 40 pins. Figure 3depicts the schematic design of the SC, indicating the used pins. This schematic
design uses the following color notation to distinguish the different pin types:
Red pins: power to 3.3 V and 5 V.
Green pins: Communication through Inter-Integrated Circuit (I2C) protocol in order to
communicate with peripherals that use this protocol.
Blue pins: Connection for the universal asynchronous receiver–transmitter (UART) for a
conventional serial port.
Black pins: Connection to ground.
Orange pins: Communication through the Serial Peripheral Interface (SPI) protocol in order to
communicate with peripherals with this protocol.
White pins: Reserved pins.
All GPIO pins: Apart from their particular function, all GPIO pins have general-purpose
Each door sensor had two wires and, due to their specification, one of these wires needed to
be connected to ground and the other one needed to be connected to some input. Pins GPIO 18,
GPIO 12, and GPIO 25 were chosen as input pins; pins GPIO 14, GPIO 20, and GPIO 30 were selected
as ground pins. These choices were mainly arbitrary, and we only considered that chosen pins
with inputs/outputs had a ground pin next to them. Because wires of door sensors could not be
Sensors 2019,19, 2552 7 of 21
directly connected to the Raspberry Pi, we connected these through a protoboard by means of jumper
wires. As one can observe in the schematic design, all ground pins were connected through jumper
wires to the positive power line of the protoboard (representing this connections with black lines).
Then, in order to ensure the connection continuity until the door sensors, a jumper wire was placed
from the power line to the central segment of the protoboard for each sensor. In the schematic design,
these connections are represented with black lines that go from the “+” column to the “A” column.
Finally, door sensors were connected to the circuit through column “E”. It was necessary to use one or
several jumper wires in this last step, depending on the distance from the protoboard to each door
sensor. Input GPIO pins were directly connected to the protoboard through the central segment,
then door sensors were connected with them through the same central segment (represented with
yellow lines).
Figure 3. Schematic design of the smart cupboard. GPIO: general-purpose input/output.
3.2. Algorithm for Measuring Memory Based on the Door Sensor Signals
The algorithm was designed to determine when a user finds an item, and when the user searches
for an item without success. One research question was: how do we know that the user had success in
searching an item inside the SC? In order to answer this question, we had to explain all possible cases
in which a user can find an object. The first case is that the user finds a certain item in the first attempt.
The second case is that the user finds a certain item in a certain number (denoted as
) of attempts,
assuming N2. The last case is the one in which the user did not find the item.
Figure 4depicts the first case. One of the rules that allows understanding this topic is that the
user usually has success in finding objects except in some cases. Nonetheless, we do not know how
many times the user has attempted to find a certain object. In this first case, the user is going to find
a certain object at the first attempt. With the open door of the SC, the user looks and searches for
the desired object. Once the user grabs the object, they close the door and at this moment, our script
takes note of one fail with the search. It seems illogical that our system would increase the fail counter,
but since we do not have another mechanism to determine exactly whether the user has grabbed the
desired item or to know whether the user has found the desired item (assuming this reduced set of
Sensors 2019,19, 2552 8 of 21
low-cost sensors), our system marks one fail. However, we kept in mind that if the user does not open
another door or the same door in a reasonable amount of time, the most probable reason is that the
user found the item. We established 10 s as reasonable amount of time, and hence once the door of
the SC was closed and 10 s passed, we estimated with high probability that the user had the desired
item. When the door is closed, besides increasing the fail counter, the “closeDoor” signal is triggered,
which starts a time counter that allows us to know whether the user opens a SC door in a reasonable
time or not. Since we are explaining the case in which an object is found at the first attempt, this time
the counter is not interrupted. In following cases we will explain what happens when this signal is
interrupted. So, when the system notices that after 10 s the user has not opened any door, it removes a
fail from the fail counter and increases a unit on the success counter.
Figure 4. Python script—First case.
The second case is more complex than the first one. We will explain the same steps as above,
but with more signals and certain features. In the second case, the user finds a certain item after
attempts. The beginning is the same as the previous case, as one can see in Figure 5, which depicts
the diagram of the second case. The user opens the door, and before they search or find anything,
the “openDoor” signal is triggered. The aim of this signal is to determine whether another door has
been opened before, and in the positive case our system cancels the count of 10 s in order to avoid
reducing the fail counter and to prevent increasing the success counter. In this way, our system counts
all fails during the process of searching for an item. Focusing again on this second case, we take for
granted that the “openDoor” signal has not been triggered, since it is the first time that the user opened
the door. Nonetheless, the action of opening the door is the trigger of the “openDoor” signal. The same
signals as in the previous case are also triggered, but we have omitted them in this description in order
to ease comprehension.
Sensors 2019,19, 2552 9 of 21
Figure 5. Python script—Second case.
When the “openDoor” signal is launched, the user starts searching for the item, then they close the
door. The fail counter increases and the “closeDoor” signal is triggered, and until this point, all steps
are the same as in the first case. Now, the user knows whether they have the correct item. If so, it is
the first case; if not, the “openDoor” signal must be interrupted. Empirical tests made by ourselves
and other users allowed us estimate that 10 s was enough time for someone to be able to open another
door of the SC; if not, again, it was assumed that the user found the correct item. Hence, in the second
case when the user thinks about what their next door choice will to be and opens it, the “openDoor”
signal is interrupted and the cycle starts again until the user opens the correct door. Until this moment
the user has always had success in searching for their object, regardless of whether it was found in
their first attempt or in attempt
. However, in the following case the user does not have any success,
since according to our point of view the user does not reach to find the required object. We kept the
third case in mind in order to manage two issues: the first is that the user searches for an object but
forgets what they were searching for. This situation gives us information about the health of the user’s
memory. The second is that in this way we can avoid that someone cheated during the experimentation
phase (i.e., if we did not control the opening time of a door and a certain object was not inside the SC,
the user would have unlimited time to think about what other compartment the object may be in).
In the experimentation phase, this case did not appear at any moment, but it is important to check for
this issue because it makes the measurement of memory loss more accurate.
Figure 6depicts the third case. The beginning is the same as aforementioned cases, only with the
difference that another signal is triggered when the door is opened by the user; this signal is called
“countTimeDoor” (we have not mentioned the signal before in order to avoid over-long explanation,
but it was also present in the previous cases). The signal’s aim is to start to count the time that an SC
door is held open. This period of time matches with period of time the user is searching an object
inside of the SC. The assigned threshold for this signal was 10 s, that is, the user has 10 s to find the
item, and if this time is surpassed the system increases the fail counter. We determined the threshold
of the “countTimeDoor” signal in the same way as we determined the 10 s in order to know whether
user was successful (i.e., through experiments made by ourselves and other users), until we were able
Sensors 2019,19, 2552 10 of 21
to determine a proper threshold. We assume that the user overcame the signal threshold, so the fail
counter increases. Furthermore, the “openDoor” signal is canceled, since if it were not so, when the
user closed the door again, our system would increase the fail counter once again. Finally, it is worth
mentioning that if the user closes the door before 10 s, the “countTimeDoor” signal is canceled in order
to avoid increasing the fail counter twice.
Figure 6. Python script—Third case.
The algorithm was written in Python, and for the sake of reproducibility, we describe how the
implementation of this algorithm handled the pins. In order to manage pins inside the script, we used
the RPi.GPIO library. There were two options to enumerate the pins. The first was GPIO mode,
in which each pin had the same number as their physical position, hence pins were enumerated
from 1 to 40. The second option was BCM (Broadcom mode), in which pins were numerated in
order to match with Broadcom chip, which was the CPU (central processing unit) of the Raspberry.
We used BCM for the implementation of the presented algorithm. The algorithm implementation
also needed to set up some pins as input pins, and this was achieved with the predefined function
GPIO.setup. Finally, the GPIO.input function provided the current state of each door sensor, true when
the two pieces of a door sensor were separated (i.e., the door was open), and false otherwise. Figure 7
presents an excerpt of the Python implementation used in the SC, showing the aforementioned
implementation details about pins. Door sensor states were used as previously described when
presenting the algorithm.
Sensors 2019,19, 2552 11 of 21
Figure 7. Implementation details about pins in the Python script of the smart cupboard.
4. Experimentation
4.1. Participants
We recruited 23 people for participation in this user study. Participants were 36.17 years old
on average (SD = 12.80) in a range from 18 to 60 years old, and had studied for 14.86 years on
average (SD = 2.88). Among the participants, only 8.69% were studying or work in computer science.
Males comprised 39.13% of participants. Participation in this experiment was voluntary and unpaid.
4.2. Procedure
In order to evaluate whether the SC is able to measure memory, several tests were conducted.
In the first test, a user was asked to observe the inside of the SC. The user had to memorize certain items
inside it in the acquisition phase. Then, the user was asked to find certain items in the retrieval phase.
A test of face–name pairs [
] was used as a control method, since this kind of test has been proven
to measure memory. The test consists of showing a list of face–name pairs, so the user memorized
them in the acquisition phase, for later selection of the name associated with each face in the retrieval
phase. The main goal was to determine whether both results were correlated, besides performing other
analyses concerning the relation of the results with the participant features.
The test was conducted in a real kitchen so that participants were familiar with the scenario.
The Raspberry was set up inside a kitchen cupboard as shown in Figure 8. The size of this cupboard
was 130 cm width
71 cm height
29.5 cm depth, and it was 150 cm above the ground. In spite of
having five compartments, we only used the bottom three in order to avoid having compartments of
different sizes.
Sensors 2019,19, 2552 12 of 21
Figure 8. Smart cupboard.
To assess the memory of participants with the SC, we selected 30 different items. All these items
were typical objects commonly found inside cupboards, and these objects were: a cup, a sweet corn can,
a chili can, an egg, a box of matches, an evaporated milk carton, a soda, a bag of breadcrumbs, a beer can,
a jar of chili peppers, a potato, a jar of lentils, a can of olives, a jar of mayonnaise, a carton of chocolate
milkshake, a can of grapes, a jar of soup cubes, a can of peaches in syrup, a can of condensed milk, salt,
a box of baking powder, a can of green peas, a milk bread, a jar of jam, a teaspoon, a jar of sausages,
honey, a can of tuna, a bag of tea, and a jar of oregano. In this experiment, each participant followed
the same process. The experimenter introduced the steps briefly to each user. In the acquisition phase,
the experimenter asked each user to memorize all the items of each compartment, and they had 30 s
per compartment—1 min and 30 s in total.
In the retrieval phase, the experimenter told the participant that he would ask them to find objects
selected in a random order from the ones inside the cupboard and previously memorized by them.
Thus, the participant had to open the compartment where they thought each required object was.
Furthermore, the experimenter indicated that only one door could be open at a time (i.e., before the
participant opened another door, they had to close the current door). Once the participant had found
the required object, grabbed it from the SC, and closed the door, the experimenter asked questions
about the item. The questions were varied and related to cooking or eating. For instance, what recipes
or dishes would you cook with this object? Or, what time of day do you usually eat this product?
Or, do you think this product is healthy and why or why not? And so on. The reasons for these
questions were to delay the retrieval phase and to increase the difficulty of the test. When users
opened the door in order to find an object, they could re-memorize where each object was (reinforcing
the initial learning), since it was unavoidable to let them see the contents again. Thus, the goals
of these distracting questions were to compensate for this aspect and to be more similar to realistic
daily conditions.
This process was repeated with ten different objects for each user. Due to the size of SC and
the amount of items, two rounds were required. In each round, each participant had an acquisition
phase in order to memorize the content of compartments concerning 15 objects, and had a retrieval
phase to sequentially find 10 objects. When the first round was finished, the experimenter asked the
user to leave the kitchen. While the participant was outside the kitchen, the experimenter set up the
second round, and then invited the participant to come into the kitchen again. Hence, users performed
20 memory retrievals about 20 different objects. In this way, this memory test had an accuracy error
margin of 5%, which is considered as appropriate in memory tests [
]. We decided that all participants
had the same conditions; hence, before starting this procedure, a list of items was selected for use by
all participants.
Sensors 2019,19, 2552 13 of 21
The list had 15 items for the first round and 15 items for the second one. The order of objects
was selected randomly, because we wanted to avoid the case where objects were organized by size or
semantic categories, in order to avoid bias in the memory measurement due to different memorization
techniques. Table 2indicates the order of objects in the SC, distributed in compartments per round.
Furthermore, another requirement of the experimentation phase was that objects could not be behind
others (i.e., all objects had to be visible from the participant’s position). The experimenter was always
with the participants, except when they were outside the kitchen. The experimenter made sure that
the participants properly followed the experimentation protocol. For instance, the experimenter was
advised to take note if any participant closed the door twice because it had not been closed with
enough strength the first time. In this case, the system would register an additional fail, so then we
could revise the logs to know what had happened.
Table 2. Order of objects in the experimentation with the smart cupboard.
Object Compartment Round Object Compartment Round
Sweet Corn Soup cubes
Chili Peaches in syrup
Egg Condensed milk
Box of Matches Salt
Evaporated milk
Baking powder
Soda Green peas
Breadcrumb Milk bread
Beer Jam
Chili peppers Teaspoon
Lentils Honey
Olives Tuna
Mayonnaise Tea
Chocolate milkshake Oregano
The next step for participants was to take a control test based on face–name pairs in order to
statistically compare these results with the SC results and to determine whether our SC is able to
measure memory. This test was similar to the common memory tests about face–name pairs in the
literature [
], and we used a short-time version that did not require several days for acquisition
and retrieval phases. In this way, each participant could do all the experimentation in the same
day, facilitating the task of recruiting unpaid volunteer participants. In this experiment, the test of
face–name pairs consisted of showing a series of 30 face–name pairs to the participant such that they
memorized these associations in the acquisition phase. Each face–name pair was presented to the
user for 6 s, and consequently the whole acquisition phase took 3 min. In the retrieval phase, each
participant had to respond to 30 questions. Each question was composed of a face image and four
name options, and the participant had to fill a form provided by the experimenter with the answers to
these questions. Figure 9shows an example of three questions. The face images have been blurred
in this article to protect the privacy of the models. The retrieval phase had no time limit, but the
experimenter instructed the participants to reply to the questions as accurately and quickly as possible,
and the reaction time was measured.
Furthermore, the participants engaged in a self-reported memory test. We performed this test
in order to compare the SC results with other memory-related variables. Since the self-assessment of
memory has proved to be relevant for evaluating memory despite other influencing factors such as
personality [
], we included this brief self-reported memory test as another control test. We selected
a short self-reported test available from the Psychology Today website (http://psychologytoday. for its brevity and
its simplicity. In this test, participants replied to seven questions with a five-point Likert scale.
The questions of this test were: (1) Do you have difficulty in remembering people’s names or phone
Sensors 2019,19, 2552 14 of 21
numbers? (2) How often do you find yourself trying to remember the location of everyday items (e.g.,
your keys, wallet, glasses, etc.)? (3) How often do you have to replace passwords (numerical or verbal)
because you’ve forgotten the original one? (4) How often do you find yourself asking questions like,
“What was I about to do next?” (5) How often do you end up arranging overlapped plans because you
forgot you had made previous plans with someone else? (6) How often do you have to ask someone to
repeat instructions or a story because you can’t remember what was said the first time? (7) How often
do you have difficulty in remembering where you parked your car? The last question was only replied
by participants who had a driving license.
The experimenter also asked participants to reply a brief demographic test to extract the
information presented in Section 4.1 when introducing the sample of participants. Once we finished
all the experimentation with all the participants, we analyzed the obtained data as described in the
next section.
Figure 9. Test of face–name pairs.
5. Results
We performed several analyses considering the memory measurements in the different methods,
the reaction time, and the age of participants. Firstly, we compared the memory measurement
results between the SC test and the face–name test, reporting the accuracies of participants in the
retrieval phase.
In order to double-check that the sensors system of the SC was working properly, the experimenter
took notes about the fails and successes of participants during the experimentation phase, and then
the notes were contrasted against the system log. The notes and SC logs and results matched perfectly.
The accuracy percentage of each participant was calculated as shown in Equation
as a measurement
of their memory:
s+f·100, (1)
where ais the accuracy percentage, sis the number of successes, and fis the number of fails.
The accuracy of each participant in the retrieval of face–name pairs was calculated in a similar
way. The experimenter checked the final test results, and the percentage was calculated as shown in
Equation (2):
n·100, (2)
is the accuracy percentage,
is the number of successes, and
is the total number of questions.
Figure 10 compares the memory accuracies of the SC and face–name pairs, and one can observe
that both measurements methods followed similar trends and shapes. Thus, there may be a correlation
between these measurement methods. In order to statistically and reliably corroborate this correlation,
we conducted a Pearson’s correlation test between the results of the two memory measurement
methods. Table 3shows the results of this correlation test. According to the Pearson correlation
coefficient [
], both memory measurement methods had a significant positive correlation. This positive
correlation was confirmed with the Kendall’s tau coefficient of 0.470 with a
-value of 0.003 and
Spearman’s rho coefficient of 0.620 with a
-value of 0.002. The correlation between the SC test and the
Sensors 2019,19, 2552 15 of 21
test of face–name pairs proves that our SC sensors system is able to measure memory, since the control
memory measurement method has already been scientifically validated.
Figure 10.
Comparison of memory measurements between the smart cupboard (SC) and the face–name
pairs test.
Table 3. Correlation between the accuracy of the SC and the accuracy of the face–name test.
Accuracy Smart
Pearson Correlation 1 0.597 **
Accuracy Smart Cupboard Sig. (2-tailed) 0.003
N23 23
Pearson Correlation 0.597 ** 1
Faces-Name Test Sig. (2-tailed) 0.003
N23 23
**. Correlation is significant at the 0.01 level (2-tailed).
Moreover, we analyzed the reaction time of participants in the SC and face–name pairs tests.
Figure 11 depicts the reaction time of each participant in both tests. The blue line represents the SC test
and the orange line represents the test of face–name pairs. In order to calculate the reaction time of the
SC test, the experimenter took note of the time spent by a participant to remember each item inside the
SC. Then, this was compared with the time in the system log in order to check that all times were correct.
Thus, 20 reaction time results were obtained for each participant. Finally, the reaction time of each
participant was calculated as the mean of all obtained times. In the face–name test, the experimenter
also measured the time spent by a participant in order to perform the test. The reaction time was
obtained from the division of the total time by the total number of questions by each participant.
According to the trend and direction of both lines in the graph, it is not easy to appreciate the similarity
in general. Nonetheless, a certain similar behavior is appreciated between participant 7 and participant
15. A similar correlation can be appreciated between these two tests, since both lines are almost parallel.
Because this observational analysis is not sufficient to determine whether there was any significant
relation between tests in reaction times, we performed another Pearson’s correlation test to statistically
determine whether there was a statistically significant correlation. Table 4depicts the result of the
correlation test. It shows that both variables were not significantly correlated. Neither Kendall’s tau
coefficient nor Spearman’s rho coefficient detected any significant correlation with respective
of 0.597 and 0.428. Thus, in these experiments, the reaction time of our SC test did not correlate with
the reaction time of the control test of face–name pairs.
Sensors 2019,19, 2552 16 of 21
Figure 11. Comparison between the SC reaction time and the reaction time of the face–name test.
Table 4. Correlation between the SC reaction time and the reaction time of the face–name test.
Reaction Time
Smart Cupboard
Reaction Time
Face-Name Test
Pearson Correlation 1 0.341
Reaction Time Smart Cupboard Sig. (2-tailed) 0.111
N23 23
Pearson Correlation 0.341 1
Reaction Time Face-Name Test Sig. (2-tailed) 0.111
N23 23
We also conducted an analysis of the relation between participant age and SC reaction times.
Figure 12 represents each participant considering these two aspects. It is worth highlighting that
we asked people to participate in these experiments, considering their age, to have both young and
aged people. In particular, there were six young participants in the 18–25 years old range and six
aged participants in the 55–60 years old range. In order to determine whether there was a statistically
significant correlation, we performed three correlation tests between SC results and age. Table 5shows
the results of the Pearson’s correlation test. One can observe that these variables were not correlated
according to this test. In addition, neither Kendall’s tau coefficient nor Spearman’s rho coefficient
detected any significant correlation, with respective
-values of 0.265 and 0.300. Thus, SC results and
age were not statistically significantly correlated in these experiments.
Figure 12. Comparison between the reaction time in the smart cupboard test and participant age.
Sensors 2019,19, 2552 17 of 21
Table 5. Correlation between the reaction time and participant age in the smart cupboard test.
Age Reaction Time
Smart Cupboard
Pearson Correlation 1 0.306
Age Sig. (2-tailed) 0.156
N23 23
Pearson Correlation 0.306 1
Reaction Time Smart Cupboard Sig. (2-tailed) 0.156
N23 23
We also performed an analysis comparing the accuracy of participants in the SC test and the
results of the self-reported memory test. Each possible answer of the self-reported memory test among
the options “almost always”, “often”, “sometimes”, “rarely”, and “almost never” were respectively
assigned the values 0, 1, 2, 3, and 4. The results of this test were standardized as a percentage calculated
with Equation (3):
Vmax , (3)
is the result of the self-reported test,
is the value of each answered question,
is the total
number of questions, and Vmax is the maximum value that a response can have (i.e., 4 in this case).
Figure 13 shows the accuracy of the SC test and this self-reported test. There were similarities
between both tests in some cases, as one can observe in the intervals between participants 1 to
4, participants 9 to 11, and participants 15 to 17. To determine the statistical significance of this
relation, we conducted a Pearson’s correlation test, and Table 6presents the results. The test indicated
that the memory results between the SC test and the self-reported one were significantly correlated.
This correlation was confirmed with the Kendall’s tau coefficient of 0.383 with a
-value of 0.014 and
Spearman’s rho coefficient of 0.451 with a
-value of 0.031. The self-reported test is a subjective test,
and personality may cause bias in the results, since this test actually measures self-perceived memory
rather than actual memory (calculated as the accuracy retrieving information previously acquired).
According to these experiments, the SC results were correlated with self-perceived memory, despite
the possible bias because of personality.
Figure 13. Comparison between the accuracy of SC and that of self-reported tests.
Sensors 2019,19, 2552 18 of 21
Table 6. Correlation between the accuracy of SC and that of self-reported tests.
Smart Cupboard Accuracy Self-Reported Test
Pearson Correlation 1 0.443 *
Smart Cupboard Sig. (2-tailed) 0.034
N23 23
Pearson Correlation 0.443 * 1
Accuracy Self-Reported Test Sig. (2-tailed) 0.034
N23 23
*. Correlation is significant at the 0.05 level (2-tailed).
6. Discussion and Conclusions
The article proposed a new mechanism of measuring memory with the SC as a novel IoT sensors
system. We presented the design of the SC as a cupboard with three sensorized doors with magnetic
door sensors connected to Internet via a Raspberry Pi 3B board with the corresponding software for
evaluating user memory.
The main goal was to have a device able to assess the memory in a familiar environment without
requiring additional effort from the user. Thus, we presented a solution in which the memory
measurement can be continuous and based on normal routine. The results based on 23 participants
in a wide age range (18 to 60 years old) showed that the accuracy of participants in finding objects
in the SC in a controlled environment was statistically significantly correlated with the accuracy of
participants in retrieving face–name associations in a validated type of memory test. The accuracy of
the SC test was also statistically significantly correlated with a self-reported memory test.
The current work attempted to find a solution that was low-cost as possible, in order to propose a
step towards a solution that can get to the market with a profitable margin, so that enterprises may
be interested and our solution can make a real impact on society. Note that the most fundamental
components of this solution were the magnetic door sensors, and in particular we used a door
sensor model that only cost $2.46. The Raspberry Pi 3B board could be easily replaced by any other
low-cost/green processing board in the market by using the same software and adapting the input
pins. Thus, a very cheap solution could potentially be developed for converting a cupboard into a
SC, so the user could install the sensor systems of their cupboard without needing to replace their
original cupboard.
In the early detection of diseases, it can be difficult to sell products to healthy people, even if
the product is cheap. We argue that as in the case of many other successful smart devices in the
market (e.g., smartphones, smartbands, and smart TVs), SCs could have multi-purpose functionalities
for successfully getting to the market. In this line, the SC could also be useful for detecting eating
patterns by classifying different kinds of food in different compartments. Eating patterns can be useful
for controlling dietary habits to reduce obesity, which is an issue for many people in countries like
the US [
]. Eating patterns could also be useful for tracking emotions by considering their known
relationship [23].
The experiments showed that the reaction time measured by the SC test did not correlate with the
reaction of the control test about face–name pairs. However, these results are not conclusive, since
the number of participants (n= 23) was not sufficient to detect medium effect sizes according to the
analysis based on statistical power performed by the G*Power 3 tool [
]. In addition, not all memory
tests need to have a correlation between reaction time and memory. In fact, strictly a memory test needs
to provide some measure that correlates with memory, which could be either accuracy or reaction time,
but not necessarily both. Thus, the proposed SC test is a reliable memory test according to the common
standards of memory tests [25].
In addition, the accuracy of the SC memory test correlated with self-perceived memory,
even though the literature supports that self-perceived memory is influenced by factors such as
personality [20], which could lead to differences between self-perceived memory and actual memory.
Sensors 2019,19, 2552 19 of 21
One limitation of the current version of the SC is that it is based on the assumption that there is
only one person using the SC to reliably measure their memory. We plan to overcome this limitation by
including an identification mechanism, which could either be (1) facial identification with a low-cost
camera following our previous work in facial authentication [
] or (2) radio-frequency identification,
which would require the user to carry a card for the identification. We will select one of these options
considering economic aspects, technological reliability, and user experience. In this manner, the SC
will determine who is using it and perform different measurements for the different family members.
In general, the inclusion of identification will allow engineers to develop SC applications for providing
customized services to the user.
In the future, we plan to conduct a study with Alzheimer’s patients over a 10-month period to
detect memory losses during the evolution of this disease, by measuring the memory of the same
persons through the study with the proposed SC-based approach. If possible, we will also enroll
people considered to probably start having Alzheimer’s disease soon, known by the analysis of genetic
information in descendants of people with Alzheimer’s. As a control group, we also plan to track the
memory evolution of a group of healthy people, which may include some of the participants of the
presented study. In this way, we plan to detect improvement opportunities and further assess whether
it is possible to track memory losses and to detect Alzheimer’s at an early stage.
Another future work is the development of an app whose main aim will be to explain to users how
to turn a normal cupboard into an SC using gamification to overcome the barrier of a possible difficult
installation. Finally, our efforts will focus on improvement of energy efficiency, since cupboards are
not used very frequently. This could be achieved by lowering the checking frequency of sensors
when users do not usually use them, based on an initial training phase, following a similar approach
to our previous one in green communications with smartbands [
]. Another option that involves
the power system is to use some energy-harvesting techniques to overcome the limitation of the
long-term use of the SC powered by batteries. The aim of harvesting techniques is to accumulate
energy from several sources that capture energy from the environment. Once the energy is accumulated,
it can be used in the SC to track user memory. Several examples of energy harvesting can be
found in the literature [
], but techniques that involve components such as micro-photovoltaic
cells, micro-thermoelectric generators [
], or indoor ambient light [
] may be the most suitable for
SCs. Furthermore, we also plan to develop an app for remotely consulting memory measurement
results from any mobile device, and to provide notifications if a family member is starting to have
significant memory losses.
Author Contributions:
Conceptualization, F.G.-L., I.G.-M. and R.A.; Data curation, F.G.-L. and I.G.-M.; Formal
analysis, F.G.-L. and I.G.-M.; Funding acquisition, I.G.-M. and R.L.; Investigation, F.G.-L., I.G.-M., R.A. and
R.L.; Methodology, F.G.-L., I.G.-M., R.A. and R.L.; Project administration, I.G.-M.; Software, F.G.-L.; Supervision,
I.G.-M., R.A. and R.L.
This work was mainly devised during the research stay of the second author in the Massachusetts
General Hospital and Harvard University, funded by “Dpto. de Innovación, Investigación y Universidad del
Gobierno de Aragón” through the program “FEDER Aragón 2014-2020 Construyendo Europa desde Aragón”
(Ref: T49_17R). This work has also been financed by the Aragonese Government and the UE through the
FEDER 2014–2020 “Construyendo Europa desde Aragón” action (Group T25_17D). We also acknowledge the
support of the projects “Collaborative Ambient Assisted living Design” (TIN2014-57028-R), “Diseño colaborativo
para la promoción del bienestar en ciudades inteligentes inclusivas” (TIN2017-88327-R), and “Red Temática
de Investigación en Ciudades Inteligentes” (TIN2016-81766-REDT) funded by the Spanish council of Science,
Innovation and Universities from the Spanish Government.
Conflicts of Interest: The authors declare no conflict of interest.
Sensors 2019,19, 2552 20 of 21
The following abbreviations are used in this manuscript:
GPIO General-Purposes Input/Output
I2C Inter-Integrated Circuit
ID Identifier
IoT Internet of Things
M2M Machine-to-Machine
SD Standard Deviation
SPI Serial Peripheral Interface
SC Smart Cupboard
TV Television
UART Universal Asynchronous Receiver-Transmitter
US United States
Prince, M.; Comas-Herrera, A.; Knapp, M.; Guerchet, M.; Karagiannidou, M. World Alzheimer Report
2016: Improving Healthcare for People Living with Dementia: Coverage, Quality and Costs Now and in the Future;
Alzheimer’s Disease International (ADI): London, UK, 2016.
Roopaei, M.; Rad, P.; Prevost, J.J. A Wearable IoT with Complex Artificial Perception Embedding for
Alzheimer Patients. In Proceedings of the 2018 World Automation Congress (WAC), Stevenson, WA, USA ,
3–6 June 2018; pp. 1–6.
Birks, J.S.; Harvey, R.J. Donepezil for dementia due to Alzheimer ’s disease. Cochrane Database Syst. Rev.
2018,6. [CrossRef] [PubMed]
Chang, C.H.; Lane, H.Y.; Lin, C.H. Brain Stimulation in Alzheimer’s Disease. Front. Psychiatry
,9, 201.
[CrossRef] [PubMed]
Prince, M.; Bryce, R.; Albanese, E.; Wimo, A.; Ribeiro, W.; Ferri, C.P. The global prevalence of dementia:
A systematic review and metaanalysis. Alzheimers Dementia 2013,9, 63–75. [CrossRef] [PubMed]
Kortuem, G.; Kawsar, F.; Sundramoorthy, V.; Fitton, D. Smart objects as building blocks for the internet of
things. IEEE Internet Comput. 2010,14, 44–51. [CrossRef]
García-Magariño, I.; González-Landero, F.; Amariglio, R.; Lloret, J. Collaboration of Smart IoT Devices
Exemplified With Smart Cupboards. IEEE Access 2019,7, 9881–9892. [CrossRef]
Nonavinakere, S.; Aldana, J.; Sisson, S.; Cruz, E.; George, K. Memory Aid Device to Improve Face-Name
Memory in Individuals with Alzheimer’s Disease. In Proceedings of the 2018 IEEE International Conference
on Healthcare Informatics (ICHI), New York, NY, USA, 4–7 June 2018; pp. 353–354.
Crema, C.; Depari, A.; Flammini, A.; Sisinni, E.; Benussi, A.; Borroni, B.; Padovani, A.
Embedded platform-based system for early detection of Alzheimer disease through transcranial magnetic
stimulation. In Proceedings of the 2018 IEEE Sensors Applications Symposium (SAS), Seoul, Korea, 12–14
March 2018.
Narendiran, A.; Nandan, M.M.; Naveen, B. Cognitive Assistance in Smart Homes to Model the Progression
of Alzheimer Disease. J. Adv. Res. Appl. Sci. 2018,5, 149–163.
Ishii, H.; Kimino, K.; Aljehani, M.; Ohe, N.; Inoue, M. An early detection system for dementia using the M2
M/IoT platform. Procedia Comput. Sci. 2016,96, 1332–1340. [CrossRef]
Kristalina, P.; Imanuddin, A.I.; Yuliana, M.; Pratiarso, A.; Astawa, I.G.P. Alzheimer Patient Tracking
System in Indoor Wireless Environment. European Scientific Journal, ESJ
,13, 327. Available online: (accessed on 3 June 2019). [CrossRef]
Chavan, S.C.; Chavan, A. Smart wearable system for fall detection in elderly people using internet of
things platform. In Proceedings of the 2017 International Conference on Intelligent Computing and Control
Systems (ICICCS), Madurai, India, 15–16 June 2017; pp. 1135–1140.
Paul, M.C.; Sarkar, S.; Rahman, M.M.; Reza, S.M.; Kaiser, M.S. Low cost and portable patient monitoring
system for e-Health services in Bangladesh. In Proceedings of the 2016 International Conference on
Computer Communication and Informatics (ICCCI), Coimbatore, India, 7–9 January 2016.
Sensors 2019,19, 2552 21 of 21
Chong, Z.H.K.; Tee, Y.X.; TOH, L.J.; Phang, S.J.; Liew, J.Y.; Queck, B.; Gottipati, S. Predicting potential
Alzheimer medical condition in elderly using IOT sensors-Case study. In Proceedings of the IRC Conference
on Science, Engineering, and Technology, Singapore, 10–11 August 2017. Available online: https://ink. (accessed on 3 June 2019).
Navarro, J.; Doctor, F.; Zamudio, V.; Iqbal, R.; Sangaiah, A.K.; Lino, C. Fuzzy adaptive cognitive stimulation
therapy generation for Alzheimer’s sufferers: Towards a pervasive dementia care monitoring platform.
Future Gener. Comput. Syst. 2018,88, 479–490. [CrossRef]
Hampstead, B.M.; Sathian, K.; Moore, A.B.; Nalisnick, C.; Stringer, A.Y. Explicit memory training leads
to improved memory for face–name pairs in patients with mild cognitive impairment: Results of a pilot
investigation. J. Int. Neuropsychol. Soc. 2008,14, 883–889. [CrossRef] [PubMed]
Hamrick, P. Declarative and procedural memory abilities as individual differences in incidental language
learning. Learn. Individ. Differ. 2015,44, 9–15. [CrossRef]
Tak, S.H.; Hong, S.H. Face-name memory in Alzheimer’s disease. Geriatr. Nurs.
,35, 290–294. [CrossRef]
Buchanan, T. Self-assessments of memory correlate with neuroticism and conscientiousness, not memory
span performance. Pers. Individ. Differ. 2017,105, 19–23. [CrossRef]
Benesty, J.; Chen, J.; Huang, Y.; Cohen, I. Pearson correlation coefficient. In Noise Reduction in Speech
Processing; Springer: New York, NY, USA, 2009.
Flegal, K.M.; Carroll, M.D.; Ogden, C.L.; Curtin, L.R. Prevalence and trends in obesity among US adults,
1999–2008. JAMA 2010,303, 235–241. [CrossRef] [PubMed]
Macht, M.; Simons, G. Emotions and eating in everyday life. Appetite
,35, 65–71. [CrossRef] [PubMed]
Faul, F.; Erdfelder, E.; Lang, A.G.; Buchner, A. G* Power 3: A flexible statistical power analysis program for
the social, behavioral, and biomedical sciences. Behav. Res. Methods
,39, 175–191. [CrossRef] [PubMed]
Ross, S.; Krukowski, R.; Putnam, S.; Adams, K. The Memory Assessment Scales in the detection of incomplete
effort in mild head injury. Clin. Neuropsychol. 2003,17, 581–591. [CrossRef] [PubMed]
Guillén-Gámez, F.D.; García-Magariño, I.; Bravo-Agapito, J.; Lacuesta, R.; Lloret, J. A proposal to improve
the authentication process in m-health environments. IEEE Access 2017,5, 22530–22544. [CrossRef]
González-Landero, F.; García-Magariño, I.; Lacuesta, R.; Lloret, J. Green communication for tracking heart
rate with smartbands. Sensors 2018,18, 2652. [CrossRef] [PubMed]
Huang, Q.; Lu, C.; Shaurette, M. Feasibility study of indoor light energy harvesting for intelligent building
environment management. In Proceedings of the International High Performance Buildings Conference,
West Lafayette, IN, USA, 12–15 July 2010. Available online:
cgi?article=1017&context=ihpbc (accesssed on 3 June 2019).
Huang, Q.; Lu, C.; Shaurette, M.; Cox, R. Environmental thermal energy scavenging powered wireless sensor
network for building monitoring. In Proceedings of the 28th International Symposium on Automation and
Robotics in Construction, Seoul, Korea, 29 June–2 July 2011; pp. 1376–1380.
Lu, C.; Raghunathan, V.; Roy, K. Efficient design of micro-scale energy harvesting systems. IEEE J. Emerg.
Sel. Top. Circuits Syst. 2011,1, 254–266. [CrossRef]
2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access
article distributed under the terms and conditions of the Creative Commons Attribution
(CC BY) license (
... Furthermore, IoT systems to monitor physical/physiological variables and the location of PwD have been proposed [17,18]. Also, data about PwD collected through IoT devices and exploited using approaches such as machine learning has been utilized to identify erratic movements [19], forgetfulness [20], sleep quality [21], language deficits [22], performance in daily activities [23], abnormal wandering patterns [24], behavioral abnormalities [25], among others. And, as a consequence, such works enable the support of the decision-making processes of caregivers and health professionals [26]. ...
... Yes [8][9][10][11][13][14][15][16][17][18][19][20][21][22][23][24][25]42,[44][45][46][47][48][49]51,[53][54][55][56][57][59][60][61][62][63][64][65][67][68][69][70][71][73][74][75][76][77][78][79][80][81]83,[85][86][87][88][89][90][92][93][94][96][97][98][99][100][101][102]104,[106][107][108][110][111][112][113][114][115][116][117][119][120][121][123][124][125][126][127] No [12,43,50,52,58,66,72,82,84,91,95,103,105,109,118,122] shows a list of the selected studies grouped by type of participant involved in the evaluation. However, it should be noted that only 28 out of the 104 selected studies presented evaluations involving participants directly, whereas the remaining selected studies evaluated their proposed IoT systems using data sets. ...
... Collecting more data [9,15,19,21,51,57,[61][62][63]74,89,92,93,102,107,113,116,121,124,127] Experimentation on real-world settings [8,9,18,53,56,64,73,76,87,89,100,110,112,114,116,117,126] Conducting further validation [11,15,21,23,42,47,56,61,63,68,87,101,111,112,120] Exploring other machine learning algorithms [17,45,60,62,68,69,85,86,88,94,99,102,106,119] Refining the model [10,24,45,46,48,57,65,75,104,106,116,123] Adding more functionalities [14,20,55,64,78,81,[96][97][98]100,111,114] Collecting different data [16,54,57,77,78,88,100,116] Recruiting more participants [17,19,23,60,74,83,117,121,125] Adjusting to the progression of patients' dementia [64,67,83,111,117] Incorporating more IoT devices [20,45,51,90,116] Long-term evaluation [10,20,23,59,101] Exploring other application domains [13,22,70,110,115] Integrating with other systems [25,80,108,112] Taking into account energy-efficiency [20,114] Using machine learning [59] Reducing the number of sensors [59] Further analyze data [44] Adding real-time capabilities [71] Correcting limitation [79] Implementing low-cost system [49] Completing the system [18] Adding intervention capabilities [83] extent, unwittingly confront patients with reality, which instead of helping them, may distress them. For instance, this could be the case of the IoT system presented in [84] that helps patients to remind faces of close people, and by doing so, it may confront patients with the fact that they were unable to remember the name or face of a close person, e.g., a family member. ...
Full-text available
Dementia is characterized by a progressive deterioration in cognitive functions and behavioral problems. Due to its importance, in the domain of Internet of Things (IoT), where physical objects are connected to the internet, a myriad of systems have been proposed to support people with dementia, their caregivers, and medical experts. However, the vast and increasing number of research efforts has led to a complex state of the art, which is in need of a methodological analysis and a characterization of its key aspects. Based on the PRISMA guidelines, this article presents a systematic review aimed at investigating the state of the art of the IoT in dementia regardless of the dementia category and/or its cause. Articles published within the period of January 2017 to November 2022 were searched in well-known scientific databases. The searches retrieved a total of 2733 records, which were narrowed down to 104 relevant studies by applying inclusion, exclusion, and quality criteria. A set of 13 research questions at the intersection of IoT and dementia were posed, which guided the analysis of the selected studies. The systematic review contributes (i) an in-depth methodological analysis of recent and relevant IoT systems in the domain of dementia; (ii) a taxonomy that identifies, characterizes, and categorizes key aspects of IoT research focused on dementia; and (iii) a series of future work directions to advance the field of IoT in the dementia domain.
... In an emergency, IoT-based detection and monitoring devices can be used to gather and store health data and save a patient's life, especially for those with specific disorders (such as cancer, diabetes, and Alzheimer's). Complex medical technologies, such as artificial hearts, joints, and organ transplants, may also be able to interact with patients on their own [21,44,49,50]. Using the IoT provides many benefits in areas such as patient health, safety, and security of pharmaceutical products by monitoring their production processes, attaching smart labels to drugs, and tracking their supply chain [44,49,50]. ...
... Complex medical technologies, such as artificial hearts, joints, and organ transplants, may also be able to interact with patients on their own [21,44,49,50]. Using the IoT provides many benefits in areas such as patient health, safety, and security of pharmaceutical products by monitoring their production processes, attaching smart labels to drugs, and tracking their supply chain [44,49,50]. ...
... Another use of IoT is the usage of smart beds connected to the Internet for determining the precise time of the patient's presence in bed [44]. In the health domain, the requirement for real-time or near-real-time data access is critical, and IoT can provide access to such data [49,50]. ...
Full-text available
Objectives: We aimed to identify and classify the Internet of Things (IoT) technologies used for Alzheimer's disease (AD)/dementia as well as the healthcare aspects addressed by these technologies and the outcomes of the IoT interventions. Methodology. We searched PubMed/MEDLINE, IEEE Explore, Web of Science, OVID, Scopus, Embase, Cochrane, and Google Scholar. In total, 13,005 papers were reviewed, 36 of which were finally selected. All the reviews were independently carried out by two researchers. In the case of any disagreement, the problem was resolved by holding a meeting and exchanging views. Due to the diversity of the reviewed studies, narrative analysis was performed. Results: Among the technologies used for the patients including radio frequency identification (RFID), near field communication (NFC), ZigBee, Bluetooth, global positioning system (GPS), sensors, and cameras, the sensors were employed in 36 studies, most of which were switch and vital sign monitoring sensors. The most common aspects of AD/dementia care monitored using these technologies were activities of daily living (ADLs) in 27 studies, followed by sleep patterns and disease diagnosis in 19 and 14 studies, respectively. Sleeping, medication, vital signs, agitation, memory, social interaction, apathy, movement, tracking, and fall were other aspects monitored by IoT. Then, their outcomes were reported. Conclusion: Using IoT for AD/dementia provides many opportunities for considering various aspects of this disease. Moreover, the ability to use various technologies for gathering patient-related data provides a comprehensive application for almost all aspects of the patients' care with high accuracy.
... Gonzalez et al. [45] designed a system that evaluates the memory of an individual. Their system includes a three-door cupboard installed with magnetic switch sensors for each door of the cupboard, a Raspberry Pi and an external battery. ...
... In the case of handheld devices, the most common signal monitored is the pattern of finger tapping and measuring response time, reaction time and rhythm [45,48,49]. These have been monitored by smartphones and simple home furniture which was made intelligent using magnetic switches and programming an Arduino board. ...
... The major advantage of such devices is that the results are quick, and their ubiquitous nature nowadays makes it a lucrative prospect for the early detection of Alzheimer's disease. Moreover, simple low-cost systems, as described in [45,50], that assess memory and visio-motor coordination of individuals can easily be implemented not only aiding in a quick diagnosis, but also these kinds of systems can help in therapy purposes as well. The repetition of such tasks strengthens their motor and cognitive abilities thus aiding in slowing down the progression of the neuronal and functional damage caused by Alzheimer's disease. ...
Full-text available
Alzheimer’s disease is a lifelong progressive neurological disorder. It is associated with high disease management and caregiver costs. Intelligent sensing systems have the capability to provide context-aware adaptive feedback. These can assist Alzheimer’s patients with, continuous monitoring, functional support and timely therapeutic interventions for whom these are of paramount importance. This review aims to present a summary of such systems reported in the extant literature for the management of Alzheimer’s disease. Four databases were searched, and 253 English language articles were identified published between the years 2015 to 2020. Through a series of filtering mechanisms, 20 articles were found suitable to be included in this review. This study gives an overview of the depth and breadth of the efficacy as well as the limitations of these intelligent systems proposed for Alzheimer’s. Results indicate two broad categories of intelligent technologies, distributed systems and self-contained devices. Distributed systems base their outcomes mostly on long-term monitoring activity patterns of individuals whereas handheld devices give quick assessments through touch, vision and voice. The review concludes by discussing the potential of these intelligent technologies for clinical practice while highlighting future considerations for improvements in the design of these solutions for Alzheimer’s disease.
... The work is based on an experiment carried out by González-Landero et al. [29], who proposed a prototype of an automated three-door cupboard with Internet of Things (IoT) technology. The system now proposed fully virtualizes this prototype and increases the cupboard with one more door (4 in total), making data collection (response times and accuracy response) in a more automated and simple way. ...
... In the training stage (study phase) the user visualizes the elements inside the cupboards. The elements are placed automatically and randomly within the shelves and the doors can be opened individually, with a waiting time for learning that has been set at 10 seconds [29]. In the evaluation stage (recall phase), the user must locate the elements requested by the application, by selecting both the door and the desired element. ...
Full-text available
Background and Objective Virtual Reality (VR) has the capacity to be used in cognitive rehabilitation interventions for diagnostic and training purposes. This technology allows the development of proposals that traditionally have been only implemented using physical elements that imply greater resources and a lesser degree of automation. This work presents an immersive virtual reality (IVR) application (the Cupboard task) for the evaluation of memory in a more ecological way and based on an activity of daily living (ADL). Methods To appraise its construct validity, we have carried out a comparative study with a traditional method of memory assessment (method of loci). To check for any association between performance and age, performance with years of education, and reaction time with age, the Pearson's correlation was used. One-way ANOVA was used to check for differences in performance by gender. We also performed a reliability analysis with a two way mixed effects model where people effects are random and measures effects are fixed. Therefore, intra-class correlation coefficient with absolute agreement was reckoned to assess the consistency or concordance of the measures made by both the method of loci and the cupboard IVR task. Results Both tasks were evaluated on a sample of 22 healthy participants who voluntarily took part in the experiment. The results obtained showed a high degree of concordance between both memory performance measures, which assumes good clinical relevance. In addition, other age-related effects were found, common to memory assessment tasks. Conclusions This work showed that it is possible to use an IVR application to successfully assess everyday memory. We have also demonstrated the potential of IVR to develop valid tests that assess memory functions reliably and efficiently and within ecologically valid contexts. The results obtained open the door to its use in clinical settings for cognitive training (and promoting cognitive health) of patients with mild cognitive impairment (MCI), severe cognitive impairment (SCI) such as Alzheimer or Dementia, etc., with full guarantees of application, although it must first be validated through a randomized control trial (RCT). The degree of usability of the Cupboard task was very high according to the test carried out by the participants.
... ing DTMC and is unable to consider variations in these tasks over time and interruptions of some activities brought on by the forgetfulness of patients. Three sensors and a Raspberry Pi were used in a smart cabinet to track the frequency of doors opened, providing a measure of the user's memory [90]. After putting the smart cupboard to the test in a controlled environment with 23 participants, a significant correlation was found between the results of the test and memory testing procedures. ...
... The application designs an activity of daily living (ADL) based on cabinets that have been previously evaluated [53,57]. The activity consists of cabinets and kitchen elements, where the user must remember the location of the elements within the cabinets. ...
Full-text available
Few works analyze the parameters inherent to immersive virtual reality (IVR) in applications for memory evaluation. Specifically, hand tracking adds to the immersion of the system, placing the user in the first person with full awareness of the position of their hands. Thus, this work addresses the influence of hand tracking in memory assessment with IVR systems. For this, an application based on activities of daily living was developed, where the user must remember the location of the elements. The data collected by the application are the accuracy of the answers and the response time; the participants are 20 healthy subjects who pass the MoCA test with an age range between 18 to 60 years of age; the application was evaluated with classic controllers and with the hand tracking of the Oculus Quest 2. After the experimentation, the participants carried out presence (PQ), usability (UMUX), and satisfaction (USEQ) tests. The results indicate no difference with statistical significance between both experiments; controller experiments have 7.08% higher accuracy and 0.27 ys. faster response time. Contrary to expectations, presence was 1.3% lower for hand tracking, and usability (0.18%) and satisfaction (1.43%) had similar results. The findings indicate no evidence to determine better conditions in the evaluation of memory in this case of IVR with hand tracking.
... Each of the subtasks consists of two phases. In the first phase (coding phase), the user was placed in front of the cupboard and was asked to memorize the elements of each of the compartments during 10 seconds [23], (40 seconds for the total of the cupboard). In the second phase (recovery phase), the participant had to locate the object required by the therapist by selecting a door. ...
Full-text available
There are conventional screening instruments for the detection of cognitive impairment, but they have a reduced ecological validity and the information they present could be biased. This study aimed at evaluating the effectiveness and usefulness of a task based on an activity of daily living (ADL) for the detection of cognitive impairment for an Alzheimer's disease (AD) population. Twenty-four participants were included in the study. The AD group (ADG) included twelve older adults (12 female) with AD (81.75±7.8 years). The Healthy group (HG) included twelve older adults (5 males, 77.7 ± 6.4 years). Both groups received a ADL-based intervention at two time frames separated 3 weeks. Cognitive functions were assessed before the interventions by using the MEC-35. The test-retest method was used to evaluate the reliability of the task, as well as the Intraclass Correlation Coefficient (ICC). The analysis of the test-retest reliability of the scores in the task indicated an excellent clinical relevance for both groups. The hypothesis of equality of the means of the scores in the two applications of the task was accepted for both the ADG and HG, respectively. The task also showed a significant high degree of association with the MEC-35 test (rho = 0.710, p = 0.010) for the ADG. Our results showed that it is possible to use an ADL-based task to assess everyday memory intended for cognitive impairments detection. In the same way, the task could be used to promote cognitive function and prevent dementia.
Conference Paper
Full-text available
Nowadays more than a billion people worldwide experience some form of disability pointing out that accessibility is a major issue that should be taken seriously into consideration. Attempting to make people’s daily habits in the kitchen area easier and more comfortable, we designed an innovative smart accessible cupboard that can identify various information about the products that are placed inside it, such as their type, quantity, location and expiration date. The Smart Kitchen Cupboard is a component of the Intelligent Kitchen aiming to support users in that space by indicating where to find a desired item, assisting in a context-sensitive manner during the cooking process and helping the overall inventory organization. Our immediate plans include planning a full-scale user evaluation in order to get useful feedback about the current design decisions so as to further improve the prototype and integrate more features.
Full-text available
In recent years, with the advent of communication technologies, healthcare sensing and remote monitoring have undergone a significant evolution to addressing almost all current e‐health challenges. In view of this, the Internet of medical things (IoMT)‐based applications are evolved. However, security and privacy are the primary concern as vast numbers of devices are connected and communicated through the wireless environment. The direct involvement of humans in IoMT‐based healthcare applications made robust and secure communication among the sensors, actuators, and patients significant. In this direction, we proposed a novel security framework for Message Queuing Transport Telemetry (MQTT) protocol based on publish/subscribe messages, which is suitable for constrained and small devices in IoMT. In this paper, we proposed a lightweight hyper elliptic curve‐multiple shared key algorithm to derive session keys in order to encrypt/decrypt health readings from the sensors connected to the patient body. The comparative analysis of performance shows that the proposed method outperforms different existing techniques in terms of computational time by reducing the computational times of broker and producer/subscriber by 0.084 and 0.0168, respectively, than the best performed existing method (Malina et al.). Finally, the security analysis shows that the proposed framework is secure against physical attacks, key control, machine‐in‐the‐middle (MITM), non‐repudiation, replay, and naming based attacks.
This book is composed by the papers written in English and accepted for presentation and discussion at The 2021 International Conference on Information Technology & Systems (ICITS 21), held at the Universidad Estatal Península de Santa Elena, in Libertad, Ecuador, between the 10th and the 12th of February 2021. ICITS is a global forum for researchers and practitioners to present and discuss recent findings and innovations, current trends, professional experiences and challenges of modern information technology and systems research, together with their technological development and applications. The main topics covered are information and knowledge management; organizational models and information systems; software and systems modelling; software systems, architectures, applications and tools; multimedia systems and applications; computer networks, mobility and pervasive systems; intelligent and decision support systems; big data analytics and applications; human–computer interaction; ethics, computers & security; health informatics; and information technologies in education.
Full-text available
The variety of smart things connected to Internet hamper the possibility of having a stand-alone solution for service-centric provisioning in Internet of Things (IoT). The different features of smart objects in processing capabilities, memory and size make it difficult for final users to learn the installation and usage of all these devices in collaboration with other IoT objects, hindering the user experience. In this context, we propose a collaboration mechanism for IoT devices based on multi-agent systems with mobile agents. This work illustrates the current approach with smart cupboards for potentially tracking memory losses. The user study revealed that users found working products of this approach usable, easy-to-learn and useful, and they agreed that the current approach could provide a high quality of experience not only in the specific case of servicecentric IoT devices for tracking memory losses but also in other domains. The learning capability by means of this approach was showed with significant reductions of reaction times and number of errors over the first and second tests with the current approach. System response times were appropriate for both continuous rendering and presenting the classification results. The usage of RAM memory was also adequate for the common actual devices.
Full-text available
The trend of using wearables for healthcare is steeply increasing nowadays, and, consequently, in the market, there are several gadgets that measure several body features. In addition, the mixed use between smartphones and wearables has motivated research like the current one. The main goal of this work is to reduce the amount of times that a certain smartband (SB) measures the heart rate (HR) in order to save energy in communications without significantly reducing the utility of the application. This work has used an SB Sony 2 for measuring heart rate, Fit API for storing data and Android for managing data. The current approach has been assessed with data from HR sensors collected for more than three months. Once all HR measures were collected, then the current approach detected hourly ranges whose heart rate were higher than normal. The hourly ranges allowed for estimating the time periods of weeks that the user could be at potential risk for measuring frequently in these (60 times per hour) ranges. Out of these ranges, the measurement frequency was lower (six times per hour). If SB measures an unusual heart rate, the app warns the user so they are aware of the risk and can act accordingly. We analyzed two cases and we conclude that energy consumption was reduced in 83.57% in communications when using training of several weeks. In addition, a prediction per day was made using data of 20 users. On average, tests obtained 63.04% of accuracy in this experimentation using the training over the data of one day for each user.
Full-text available
In this paper, we present a novel system for cognitive stimulation therapy to progressively assess cognitive impairment and emotional well-being of dementia patients in social care settings. The system assesses patients interactions and computes performance scores for different areas of cognitive stimulation. Patient interactions are initially classified into predefined performance categories through clustering of a sampled population. New personalized stimulation plans tailored to match the patient's changing level of impairment are generated automatically through a set of fuzzy rule based systems using quantitative attributes and the overall scores of patients interactions. Therapists can redefine, evaluate and adjust the rules governing difficulty and activity levels for different stimulation areas to fine tune generated activity plans. The system can also be combined with an Internet of Things (IoT) enabled patient dialogue system for determining the affective state of participants during therapy sessions that could be used as a pervasive condition monitoring platform. Experiments consisting of therapy sessions of patients interacting with the system were performed in which the activity plans were automatically generated. Initial results showed that the system outputs were in agreement with the therapists own assessment in most of the stimulation areas. Simulation experiments were also conducted to analyse the system performance over multiple sessions. The results suggest that the system is able to adapt therapy plans overtime in response to changing levels of impairment/performance while supporting therapists to tune and evaluate therapy plans more effectively.
Full-text available
Brain stimulation techniques can modulate cognitive functions in many neuropsychiatric diseases. Pilot studies have shown promising effects of brain stimulations on Alzheimer's disease (AD). Brain stimulations can be categorized into non-invasive brain stimulation (NIBS) and invasive brain stimulation (IBS). IBS includes deep brain stimulation (DBS), and invasive vagus nerve stimulation (VNS), whereas NIBS includes transcranial magnetic stimulation (TMS), transcranial direct current stimulation (tDCS), transcranial alternating current stimulation (tACS), electroconvulsive treatment (ECT), magnetic seizure therapy (MST), cranial electrostimulation (CES), and non-invasive VNS. We reviewed the cutting-edge research on these brain stimulation techniques and discussed their therapeutic effects on AD. Both IBS and NIBS may have potential to be developed as novel treatments for AD; however, mixed findings may result from different study designs, patients selection, population, or samples sizes. Therefore, the efficacy of NIBS and IBS in AD remains uncertain, and needs to be further investigated. Moreover, more standardized study designs with larger sample sizes and longitudinal follow-up are warranted for establishing a structural guide for future studies and clinical application.
Conference Paper
Full-text available
Alzheimer Disease (AD) represents one of the main causes of neurodegenerative dementia in the population aged 60+. This fact, together with the growing number of elderly forecasted for the next decades, will increase its relevance in the medium-term future. In this view, numerous disease-modifying treatments are being developed, which have been shown to be effective only if administered early in the disease course. Thus, the identification of the dementia typology afflicting a patient is of primary importance. Despite several techniques have been proposed in the last decades, it is still challenging to distinguish AD from other neurodegenerative dementias. Transcranial Magnetic Stimulation (TMS) techniques show good results; moreover, they are non-invasive, easy to apply and not time consuming. TMS-based techniques are usually implemented by means of ad hoc clinical equipment, purposely designed and configured to implement a specific diagnostic protocol. Devices are often expensive, and they require specialized personnel to implement the setup and to manage the measuring process. This work proposes a new architecture for the application of TMS which can provide a high degree of versatility, being suitable for a large number of diagnostic protocols, as well as easiness of use/configuration, being based on a simple web interface. Moreover, test results are promptly available for medical interpretation, thus speeding up the diagnostic process.
Background: Alzheimer's disease is the most common cause of dementia in older people. One approach to symptomatic treatment of Alzheimer's disease is to enhance cholinergic neurotransmission in the brain by blocking the action of the enzyme responsible for the breakdown of the neurotransmitter acetylcholine. This can be done by a group of drugs known as cholinesterase inhibitors. Donepezil is a cholinesterase inhibitor.This review is an updated version of a review first published in 1998. Objectives: To assess the clinical efficacy and safety of donepezil in people with mild, moderate or severe dementia due to Alzheimer's disease; to compare the efficacy and safety of different doses of donepezil; and to assess the effect of donepezil on healthcare resource use and costs. Search methods: We searched Cochrane Dementia and Cognitive Improvement's Specialized Register, MEDLINE, Embase, PsycINFO and a number of other sources on 20 May 2017 to ensure that the search was as comprehensive and up-to-date as possible. In addition, we contacted members of the Donepezil Study Group and Eisai Inc. Selection criteria: We included all double-blind, randomised controlled trials in which treatment with donepezil was administered to people with mild, moderate or severe dementia due to Alzheimer's disease for 12 weeks or more and its effects compared with those of placebo in a parallel group of patients, or where two different doses of donepezil were compared. Data collection and analysis: One reviewer (JSB) extracted data on cognitive function, activities of daily living, behavioural symptoms, global clinical state, quality of life, adverse events, deaths and healthcare resource costs. Where appropriate and possible, we estimated pooled treatment effects. We used GRADE methods to assess the quality of the evidence for each outcome. Main results: Thirty studies involving 8257 participants met the inclusion criteria of the review, of which 28 studies reported results in sufficient detail for the meta-analyses. Most studies were of six months' duration or less. Only one small trial lasted 52 weeks. The studies tested mainly donepezil capsules at a dose of 5 mg/day or 10 mg/day. Two studies tested a slow-release oral formulation that delivered 23 mg/day. Participants in 21 studies had mild to moderate disease, in five studies moderate to severe, and in four severe disease. Seventeen studies were industry funded or sponsored, four studies were funded independently of industry and for nine studies there was no information on source of funding.Our main analysis compared the safety and efficacy of donepezil 10 mg/day with placebo at 24 to 26 weeks of treatment. Thirteen studies contributed data from 3396 participants to this analysis. Eleven of these studies were multicentre studies. Seven studies recruited patients with mild to moderate Alzheimer's disease, two with moderate to severe, and four with severe Alzheimer's disease, with a mean age of about 75 years. Almost all evidence was of moderate quality, downgraded due to study limitations.After 26 weeks of treatment, donepezil compared with placebo was associated with better outcomes for cognitive function measured with the Alzheimer's Disease Assessment Scale-Cognitive (ADAS-Cog, range 0 to 70) (mean difference (MD) -2.67, 95% confidence interval (CI) -3.31 to -2.02, 1130 participants, 5 studies), the Mini-Mental State Examination (MMSE) score (MD 1.05, 95% CI 0.73 to 1.37, 1757 participants, 7 studies) and the Severe Impairment Battery (SIB, range 0 to 100) (MD 5.92, 95% CI 4.53 to 7.31, 1348 participants, 5 studies). Donepezil was also associated with better function measured with the Alzheimer's Disease Cooperative Study activities of daily living score for severe Alzheimer's disease (ADCS-ADL-sev) (MD 1.03, 95% CI 0.21 to 1.85, 733 participants, 3 studies). A higher proportion of participants treated with donepezil experienced improvement on the clinician-rated global impression of change scale (odds ratio (OR) 1.92, 95% CI 1.54 to 2.39, 1674 participants, 6 studies). There was no difference between donepezil and placebo for behavioural symptoms measured by the Neuropsychiatric Inventory (NPI) (MD -1.62, 95% CI -3.43 to 0.19, 1035 participants, 4 studies) or by the Behavioural Pathology in Alzheimer's Disease (BEHAVE-AD) scale (MD 0.4, 95% CI -1.28 to 2.08, 194 participants, 1 study). There was also no difference between donepezil and placebo for Quality of Life (QoL) (MD -2.79, 95% CI -8.15 to 2.56, 815 participants, 2 studies).Participants receiving donepezil were more likely to withdraw from the studies before the end of treatment (24% versus 20%, OR 1.25, 95% CI 1.05 to 1.50, 2846 participants, 12 studies) or to experience an adverse event during the studies (72% vs 65%, OR 1.59, 95% 1.31 to 1.95, 2500 participants, 10 studies).There was no evidence of a difference between donepezil and placebo for patient total healthcare resource utilisation.Three studies compared donepezil 10 mg/day to donepezil 5 mg/day over 26 weeks. The 5 mg dose was associated with slightly worse cognitive function on the ADAS-Cog, but not on the MMSE or SIB, with slightly better QoL and with fewer adverse events and withdrawals from treatment. Two studies compared donepezil 10 mg/day to donepezil 23 mg/day. There were no differences on efficacy outcomes, but fewer participants on 10 mg/day experienced adverse events or withdrew from treatment. Authors' conclusions: There is moderate-quality evidence that people with mild, moderate or severe dementia due to Alzheimer's disease treated for periods of 12 or 24 weeks with donepezil experience small benefits in cognitive function, activities of daily living and clinician-rated global clinical state. There is some evidence that use of donepezil is neither more nor less expensive compared with placebo when assessing total healthcare resource costs. Benefits on 23 mg/day were no greater than on 10 mg/day, and benefits on the 10 mg/day dose were marginally larger than on the 5 mg/day dose, but the rates of withdrawal and of adverse events before end of treatment were higher the higher the dose.
Alzheimer's disease is a disease of the nerves that are irreversible, resulting in memory impairment. This condition resulted in Alzheimer's patients easily lost because they forget the existence. In this research, we designed a tracking system for Alzheimer's patients in a hospital environment, incorporating Kalman method to estimate the position of the patient. As known Received Signal Strength Indicator value is strongly influenced by environmental conditions that lead to the acquisition of position estimation is inaccurate. From the test results showed that the optimal Kalman estimated value obtained when the value of R = 0:01 and Q = 0.1 with the average percentage of error only 7.01 % of the actual patient position. The test results with various data variations also indicate the reliability of the Kalman method, because of the average estimated position approach the actual patient position.