To read the full-text of this research, you can request a copy directly from the authors.
... Las técnicas de diagnóstico por imágenes del tórax  se han utilizado como una herramienta diagnóstica en el departamento de emergencias porque pueden revelar características relacionadas con la afectación en pulmones de covid-19. La imagen del tórax o radiografía (Rx) simple de tórax es una técnica sencilla, segura y de amplio uso en la valoración inicial de la covid-19. ...
... El uso de algoritmos de aprendizaje automático para el diagnóstico de enfermedades puede reducir los tiempos de consulta e incluso evitar la consulta con el especialista. Son un área del conocimiento donde convergen diferentes técnicas, en particular las RN, los Árboles de Decisión (AD) y las redes bayesianas teniendo un gran desarrollo e impacto en la medicina (Montero Rodríguez et al., 2019). El modelo para la predicción de supervivencia en pacientes con trauma basado AA presentado por Rau et al. (2019), demostró que tanto los modelos de regresión lineal, máquina de vectores de soporte, RN y puntaje de gravedad de lesiones y traumas exhibieron una alta precisión similar en la predicción de la supervivencia de los pacientes con trauma. ...
Uno de los campos con mayor crecimiento en los últimos años ha sido la inteligencia artificial, la cual se divide en subcampos como el aprendizaje automático, el cual provee técnicas y algoritmos para que los sistemas puedan aprender y mejorar de forma automática. El objetivo del proyecto fue desarrollar una aplicación médica cuya función radica en optimizar y facilitar la toma de decisiones referente a la atención brindada a pacientes de trauma craneoncefálico, por medio de un modelo predictivo que indique la probabilidad de muerte de estos pacientes. Se aplicó la metodología de programación extrema (XP), empleando técnicas de aprendizaje automático tomando como base el conjunto de datos CRASH-2, la cuál cuenta con 20,207 registros de pacientes aleatorizados de trauma craneoencefálico. Las redes neuronales artificiales fueron utilizadas para construir el modelo predictivo de supervivencia. La red neuronal con la arquitectura 6-(8-14-2)-1 alcanzó una exactitud del 76%, sensibilidad del 72.9% y especificidad del 94.1% sobre el conjunto de datos de prueba; demostrando una capacidad de discriminación prometedora con buena adaptación a la validación interna.
... The performance was assessed by comparison of pretrained models using various metrics. The performance of pre-trained learning models are usually performed through evaluating the test dataset . ...
Medicine faces the challenge of acquiring, analyzing and applying knowledge to solve complex clinical problems. The accelerated development of technology has allowed the storage and processing of large volumes of information composed of different types of data, which are not always as accurate and complete as necessary. Therefore, the pre-processing of data is a step prior to obtaining quality data and from them perform disease diagnoses using AI techniques, in particular the Bayesian networks.
Introducción: Por tratarse de una tarea altamente compleja y de importancia clínica, el diagnóstico del síndrome coronario agudo se presta para su exploración por medio de modelado mediante sistemas inteligentes.
Objetivo: desarrollar un sistema multiagente que ensamble las decisiones de varias redes neuronales para el diagnóstico del dolor torácico enfocado a los síndromes coronarios agudos.
Metodología: estudio de pruebas diagnósticas en el que se entrenan un conjunto de redes neuronales con una precisión cercana al 70%, que luego son ensambladas mediante tres sistemas de votación para luego adicionar el resultado de redes especiales en poblaciones particulares y seleccionar la mejor configuración que hará parte de un sistema multiagente para el diagnóstico del dolor torácico.
Resultados: Se generaron 84 redes con precisión promedio del 72% en pruebas; al ensamblarse aumentan dicha precisión hasta llegar a un máximo del 84% que tras la adición de los grupos especiales alcanza el 89%. Se escoge una conformación que brinda una sensibilidad del 96% con una especificidad del 77%, con valores predictivos positivo y negativo de 87 y 93% respectivamente para el diagnóstico de síndrome coronario agudo.
Conclusiones: Es posible desarrollar una herramienta para el diagnóstico automático del síndrome coronario agudo a partir de un sistema multiagente que ensamble la disposición tomada por un conjunto de redes neuronales artificiales, cuyo rendimiento permite su consideración para su implementación dentro de un sistema de soporte a las decisiones clínicas.
Malignant pleural mesothelioma (MPM) is a rare cancer with a heterogeneous prognosis. Prognostic models are not widely utilised clinically. Classification and regression tree (CART) analysis examines the interaction of multiple variables with a given outcome.
Between 2005-2014, all cases with pathologically confirmed MPM had routinely available histological, clinical and laboratory characteristics recorded. CART analysis was performed using 29 variables with 18-month survival as the dependent variable. Risk groups were refined according to survival and clinical characteristics. The model was then tested on an external international cohort.
482 cases were included in the derivation cohort, median survival was 12.6 months, median age 69 years. The model defined four risk groups with clear survival differences (p<0.0001). The strongest predictive variable was the presence of weight loss. The group with the best survival at 18 months (86.7% alive: 'risk group 1', median survival 34.0 months) had no weight loss, haemoglobin >153 g/L and serum albumin >43 g/L. The group with the worst survival (0% alive: 'risk group 4d', median survival 7.5 months) had weight loss, performance score 0-1 and sarcomatoid histology. The C-statistic for the model was 0.761, sensitivity 94.5%. Validation on 174 external cases confirmed the model's ability to discriminate between risk groups on an alternative dataset with fair performance (C-statistic 0.68).
We have developed and validated a simple, clinically relevant model to reliably discriminate cases at high and lower risk of death using routinely available variables from the time of diagnosis in unselected populations of MPM patients.
Comprehensive introduction to the neural network models currently under intensive study for computational applications. It also provides coverage of neural network applications in a variety of problems of both theoretical and practical interest.
To provide an improved method for the identification and analysis of brain tumors in MRI scans using a semi-automated computational approach, that has the potential to provide a more objective, precise and quantitatively rigorous analysis, compared to human visual analysis.
Self-Organizing Maps (SOM) is an unsupervised, exploratory data analysis tool, which can automatically domain an image into selfsimilar regions or clusters, based on measures of similarity. It can be used to perform image-domain of brain tissue on MR images, without prior knowledge.
We used SOM to analyze T1, T2 and FLAIR acquisitions from two MRI machines in our service from 14 patients with brain tumors confirmed by biopsies - three lymphomas, six glioblastomas, one meningioma, one ganglioglioma, two oligoastrocytomas and one astrocytoma. The SOM software was used to analyze the data from the three image acquisitions from each patient and generated a self-organized map for each containing 25 clusters.
Damaged tissue was separated from the normal tissue using the SOM technique. Furthermore, in some cases it allowed to separate different areas from within the tumor - like edema/peritumoral infiltration and necrosis. In lesions with less precise boundaries in FLAIR, the estimated damaged tissue area in the resulting map appears bigger.
Our results showed that SOM has the potential to be a powerful MR imaging analysis technique for the assessment of brain tumors.
Liver cirrhosis has acquired a great importance on both national and global levels due to the growing number of ill persons and nevertheless to serious complication associated to it. Worldwide liver cirrhosis (liver cirrhosis) represents the tenth leading cause of death according to recent statistical data reported by the World Health Organization: The prior concern of medical science for is to establish an effective diagnostic algorithm for liver cirrhosis and to implement therapeutic protocols in order to achieve an adequate management of complications. Time based the correct diagnose of liver cirrhosis can be essential in order to prevent further liver damage. That is translated in according the ill patient a real chance for transplantation and preventing decompensation risk factors for this condition. The main goal of this paper is to design a noninvasive method based on an artificial neural network model that will serve to diagnose liver cirrhosis patients by using only laboratory data. The prospective study included patients with various etiologies liver cirrhosis hospitalized or treated in the Gastroenterology Clinic of the Emergency Hospital 'St. Andrew' from Galati which have been monitored every 3 months for one year.
To use computational intelligence models for the classification and identification of endophenotype (relationships between phenotype and genetic markers) in patients with rheumatoid arthritis and healthy controls from genetic information, primarily the DRB1 HLA (human leukocyte antigen) and the shared epitope theory. This refers to the association between rheumatoid arthritis and the HLA-DRB1 alleles mainly containing amino acid common motif sequences QKRAA, RRRAA, QRRAA or at positions 70 to 74 DRB1 chain, which have been associated to susceptibility of this disease.
Development of reliable medical decision support systems has been the subject of many studies among which Artificial Neural Networks (ANNs) gained increasing popularity and gave promising results. However, wider application of ANNs in clinical practice remains limited due to the lack of a standard and intuitive procedure for their configuration and evaluation which is traditionally a slow process depending on human experts. The principal contribution of this study is a novel procedure for obtaining ANN predictive models with high performances. In order to reach those considerations with minimal user effort, optimal configuration of ANN was performed automatically by Genetic Algorithms (GA). The only two user dependent tasks were selecting data (input and output variables) and evaluation of ANN threshold probability with respect to the Regret Theory (RT). The goal of the GA optimization was reaching the best prognostic performances relevant for clinicians: correctness, discrimination and calibration. After optimally configuring ANNs with respect to these criteria, the clinical usefulness was evaluated by the RT Decision Curve Analysis. The method is initially proposed for the prediction of advanced bladder cancer (BC) in patients undergoing radical cystectomy, due to the fact that it is clinically relevant problem with profound influence on health care. Testing on the data of the ten years cohort study, which included 183 evaluable patients, showed that soft max activation functions and good calibration were the most important for obtaining reliable BC predictive models for the given dataset. Extensive analysis and comparison with the solutions commonly used in literature showed that better prognostic performances were achieved while user-dependency was significantly reduced. It is concluded that presented procedure represents a suitable, robust and user-friendly framework with potential to have wide applications and influence in further development of health care decision support systems.
The aim of the present study was to evaluate the performance of 2 different multivariate statistical methods and artificial neural networks (ANNs) in predicting the mortality of hemorrhagic and ischemic patients within the first 10 days after stroke.
The multilayer perceptron (MLP) ANN model and multivariate statistical methods (multivariate discriminant analysis [MDA] and logistic regression analysis [LRA]) have been used to predict acute stroke mortality. The data of total 570 patients (230 hemorrhagic and 340 ischemic stroke), who were admitted to the hospital within the first 24 hours after stroke onset, have been used to develop prediction models. The factors affecting the prognosis were used as inputs for prediction models. Survival or death status of the patients was taken as output of the models.
For the MLP method, the accuracies were 99.9% in a training data set and 80.9% in a testing data set for the hemorrhagic group, whereas 97.8% and 75.9% for the ischemic group, respectively. For the MDA method, the training and testing performances were 89.8%, 87.8% and 80.6%, 79.7% for hemorrhagic and ischemic groups, respectively. For the LRA method, the training and testing performances for the hemorrhagic group were 89.7% and 86.1%, and for the ischemic group were 81.7% and 80.9%, respectively.
Training and test performances yielded different results for ischemic and hemorrhagic groups. MLP method was most successful for the training phase, whereas LRA and MDA methods were successful for the test phase. In the hemorrhagic group, higher prediction performances were achieved for both training and testing phases.
The aim of this research is to propose a new neural network based method for medical image segmentation. Firstly, a modified self-organizing map (SOM) network, named moving average SOM (MA-SOM), is utilized to segment medical images. After the initial segmentation stage, a merging process is designed to connect the objects of a joint cluster together. A two-dimensional (2D) discrete wavelet transform (DWT) is used to build the input feature space of the network. The experimental results show that MA-SOM is robust to noise and it determines the input image pattern properly. The segmentation results of breast ultrasound images (BUS) demonstrate that there is a significant correlation between the tumor region selected by a physician and the tumor region segmented by our proposed method. In addition, the proposed method segments X-ray computerized tomography (CT) and magnetic resonance (MR) head images much better than the incremental supervised neural network (ISNN) and SOM-based methods.
OBJECTIVE: To construct an artificial neural network (ANN) model to predict survival after liver resection for colorectal cancer (CRC) metastases. BACKGROUND: CRC liver metastases are fatal if untreated and resection can possibly be curative. Predictive models stratify patients into risk categories to predict prognosis and select those who can benefit from aggressive multidisciplinary treatment and intensive follow-up. Standard linear models assume proportional hazards, whereas more flexible non-linear survival models based on ANNs may better predict individual long-term survival. METHODS: Clinicopathological and perioperative data on patients who underwent liver resection for CRC metastases between 1994 and 2009 were studied retrospectively. A five-fold cross-validated ANN model was constructed. Risk variables were ranked and minimised through calibrated ANNs. Time dependent hazard ratio (HR) was calculated using the ANN. Performance of the ANN model and Cox regression were analysed using Harrell's C-index. RESULTS: 241 patients with a median age of 66 years were included. There were no perioperative deaths and median survival was 56 months. Of 28 potential risk variables, the ANN selected six: age, preoperative chemotherapy, size of largest metastasis, haemorrhagic complications, preoperative CEA-level and number of metastases. The C-index was 0.72 for the ANN model and 0.66 for Cox regression. CONCLUSION: For the first time ANNs were used to successfully predict individual long-term survival for patients following liver resection for CRC metastases. In the future, more complex prognostic factors can be incorporated into the ANN model to increase its predictive ability.
Artificial neural networks (ANNs) are nonlinear pattern recognition techniques that can be used as a tool in medical decision making. The objective of this study was to develop an ANN model for predicting survival in patients with pancreatic ductal adenocarcinoma (PDAC).
A flexible nonlinear survival model based on ANNs was designed by using clinical and histopathological data from 84 patients who underwent resection for PDAC.
Seven of 33 potential risk variables were selected to construct the ANN, including lymph node metastasis, differentiation, body mass index, age, resection margin status, peritumoral inflammation, and American Society of Anesthesiologists grade. Three variables (ie, lymph node metastasis, leukocyte count, and tumor location) were significant according to Cox regression analysis. Harrell's concordance index for the ANN model was .79, and for Cox regression it was .67.
For the first time, ANNs have been used to successfully predict individual long-term survival for patients after radical surgery for PDAC.
A unified view of metaheuristics. This book provides a complete background on metaheuristics and shows readers how to design and implement efficient algorithms to solve complex optimization problems across a diverse range of applications, from networking and bioinformatics to engineering design, routing, and scheduling. It presents the main design questions for all families of metaheuristics and clearly illustrates how to implement the algorithms under a software framework to reuse both the design and code. Throughout the book, the key search components of metaheuristics are considered as a toolbox for: Designing efficient metaheuristics (e.g. local search, tabu search, simulated annealing, evolutionary algorithms, particle swarm optimization, scatter search, ant colonies, bee colonies, artificial immune systems) for optimization problems. Designing efficient metaheuristics for multi-objective optimization problems. Designing hybrid, parallel, and distributed metaheuristics. Implementing metaheuristics on sequential and parallel machines. Using many case studies and treating design and implementation independently, this book gives readers the skills necessary to solve large-scale optimization problems quickly and efficiently. It is a valuable reference for practicing engineers and researchers from diverse areas dealing with optimization or machine learning; and graduate students in computer science, operations research, control, engineering, business and management, and applied mathematics.
In this work we present a method based on partial decision trees and association rules for the prediction of Parkinson's disease (PD) symptoms. The proposed method is part of the PERFORM system. PERFORM is used for the treatment of PD patients and even advocate specific combinations of medications. The approach presented in this paper is included in the data miner module of PERFORM. A patient performs some initial examinations and the module predicts the future occurrence of the symptoms based on the initial examinations and medications taken. Using the method, the expert can prescribe specific medications that will not cause, or postpone the appearance of specific symptoms to the patient. The approach employed is able to provide interpretation for the predictions made, by providing rules. The models have been developed and evaluated using real patient's data and the respective results are reported. Another functionality of the data miner module is the extraction of rules through a user friendly interface using association rule mining algorithms. These rules can be used for the prediction analysis of patient's reaction to certain treatment plans. The accuracy of the symptoms' prediction ranges from 57.1 to 77.4%, depending on the symptom.
Right ventricular (RV) failure is a significant complication after implantation of a left ventricular assist device (LVAD). It is therefore important to identify patients at risk a priori. However, prognostic models derived from multivariate analyses have had limited predictive power.
This study retrospectively analyzed the records of 183 LVAD recipients between May 1996 and October 2009; of these, 27 later required a RVAD (RVAD(+)) and 156 remained on LVAD only (RVAD(-)) until transplant or death. A decision tree model was constructed to represent combinatorial non-linear relationships of the pre-operative data that are predictive of the need for RVAD support.
An optimal set of 8 pre-operative variables were identified: transpulmonary gradient, age, right atrial pressure, international normalized ratio, heart rate, white blood cell count, alanine aminotransferase, and the number of inotropic agents. The resultant decision tree, which consisted of 28 branches and 15 leaves, identified RVAD(+) patients with 85% sensitivity, RVAD(-) patients with 83% specificity, and exhibited an area under the receiver operating characteristic curve of 0.87.
The decision tree model developed in this study exhibited several advantages compared with existing risk scores. Quantitatively, it provided improved prognosis of RV support by encoding the non-linear, synergic interactions among pre-operative variables. Because of its intuitive structure, it more closely mimics clinical reasoning and therefore can be more readily interpreted. Further development with additional multicenter, longitudinal data may provide a valuable prognostic tool for triage of LVAD therapy and, potentially, improve outcomes.
In this study, an artificial neural network (ANN) was developed to determine whether patients have breast cancer or not. Whether patients have cancer or not and if they have its type can be determined by using ANN and BI-RADS evaluation and based on the age of the patient, mass shape, mass border and mass density. Though this system cannot diagnose cancer conclusively, it helps physicians in deciding whether a biopsy is required by providing information about whether the patient has breast cancer or not. Data obtained from 800 patients who were diagnosed with cancer definitively through biopsy. The definitive diagnosis corresponding to each patient and the data from ANN model results were investigated using Confusion matrix and ROC analyses. In the test data of the ANN model that was implemented as a result of these analyses, disease prediction rate was 90.5% and the health ratio was 80.9%. It is seen from these high predictive values that the ANN model is fast, reliable and without any risks and therefore can be of great help to physicians.
To assess the performance of general severity systems (Acute Physiology and Chronic Health Evaluation [APACHE] II, Simplified Acute Physiology Score [SAPS] II, and Mortality Probability Models [MPM] II) for head trauma patients and to compare these systems with the Glasgow Coma Score, in order to obtain a good estimate of severity of illness and probability of hospital mortality.
Adult medical and surgical intensive care units in 12 European and North American countries.
Patients (n = 401) who were diagnosed with head trauma (with/without multiple trauma), leading to intensive care unit admission, and who were not brain dead at the time of arrival.
Statistical analysis to assess the performance of general severity systems.
Vital status at the time of hospital discharge was the outcome measure. Performance of the severity systems (SAPS II, MPM II0 [MPM at admission], MPM II24 [MPM at 24 hrs], and APACHE II) was assessed by evaluating calibration and discrimination. Logistic regression was used to convert the Glasgow Coma Score into a probability of death. The MPM II system (either MPM II0 or MPM 1124) provided an adequate estimation of the mortality experience in patients with head trauma. SAPS II and APACHE II systems did not calibrate well, although they showed high discrimination (area under the receiver operating characteristic curve 0.95 for SAPS II, 0.94 for APACHE II, and 0.90 for MPM II0 and MPM II24). The logistic regression model containing the Glasgow Coma Score as an independent variable and developed in this group of patients was not as well calibrated as MPM II. The discrimination of this model was very high, in the range observed for the APACHE II, SAPS II, and MPM II systems.
The MPM II system performs better than APACHE II, SAPS II, and Glasgow Coma Score for head trauma patients. If our results are supported by other studies, MPM II would be an appropriate tool to assess severity of illness in head trauma patients, with applications to clinical practice and clinical research.
Estudio y selección de las técnicas de Inteligencia Artificial para el diagnóstico de enfermedades
N Estrada Senti
V Febles Estrada
Gonzalez Benitez, N., Estrada Senti, V., Febles Estrada, A.: Estudio y selección
de las técnicas de Inteligencia Artificial para el diagnóstico de enfermedades. Rev
Ciencias Médicas, Pinar del Río 22(3), 131-141 (2018)
Aplicación De Métodos De Inteligencia Artificial en elÁrea Médica
N V Ramírez Pérez
A M De Anda Tovar
Ramírez Pérez, N.V., Laguna Estrada, M., De Anda Tovar, A.M.: Aplicación De
Métodos De Inteligencia Artificial en elÁrea Médica. Pistas Educativas 35(111),
Diagnosis of Heart Disease by Using a Radial Basis Function Network classification technique on patients' medical records
Alsalamah, M., Amin, S., Halloran, J.: Diagnosis of Heart Disease by Using a Radial
Basis Function Network classification technique on patients' medical records. In RF
and Wireless Technologies for Biomedical and Healthcare Applications (IMWS-Bio), IEEE MTT-S International Microwave Workshop Series on, pp. 1-4 (2014)
Breast cancer diagnosis based on mammary thermography and extreme
Araujo, A.: Breast cancer diagnosis based on mammary thermography and extreme. Learning Machines (2018)
Árboles de decisiones en el diagnóstico de enfermedades cardiovasculares
G R Solarte Martínez
J A Soto Mejía
Solarte Martínez, G.R., Soto Mejía, J.A.:Árboles de decisiones en el diagnóstico
de enfermedades cardiovasculares. Scientia Et Technica XVI (2011)
Aplicación de Redes Bayesianas en el Modelado de un Sistema Experto de Triaje en Servicios de Urgencias Médicas
M M Abad-Grau
Abad-Grau, M.M.: Aplicación de Redes Bayesianas en el Modelado de un Sistema
Experto de Triaje en Servicios de Urgencias Médicas. En: IX Workshop de
Investigadores en Ciencias de la Computación, pp. 43-47 (2007)