A new approach for microscopic diagnosis of malaria parasites in thick blood smears using pre-trained deep learning models

Abstract

One of the deadly endemic diseases in sub-Saharan Africa is malaria. Its prevalence is promoted by lack of sufficient expertise to carry out accurate and timely diagnosis using the standard microscopy method. Where lab technicians are available, the results are usually subjective due to variations in expert judgement. To address this challenge, prompt interventions to improve disease control are needed. The emerging technologies of machine learning that can learn complex image patterns have accelerated research in medical image analysis. In this study, on a dataset of thick blood smear images, we evaluate and compare performance of three pre-trained deep learning architectures namely; faster regional convolutional neural network (faster R-CNN), single-shot multi-box detector (SSD) and RetinaNet through a Tensorflow object detection API. Data augmentation method was applied to optimise performance of the meta architectures. The possibility for mobile phone detector deployment was also investigated. The results revealed that faster R-CNN was the best trained model with a mean average precision of over 0.94 and SSD, was the best model for mobile deployment. We therefore deduce that faster R-CNN is best suited for obtaining high rates of accuracy in malaria detection while SDD is best suited for mobile deployment.

Full text

Vol.:(0123456789)
SN Applied Sciences (2020) 2:1255 | https://doi.org/10.1007/s42452-020-3000-0
Research Article
A new approach formicroscopic diagnosis ofmalaria parasites inthick
blood smears using pre‑trained deep learning models
RoseNakasi1· ErnestMwebaze2· AminahZawedde1· JeremyTusubira1· BenjaminAkera1· GilbertMaiga1
Received: 6 December 2019 / Accepted: 1 June 2020 / Published online: 20 June 2020
© Springer Nature Switzerland AG 2020
Abstract
One of the deadly endemic diseases in sub-Saharan Africa is malaria. Its prevalence is promoted by lack of sucient
expertise to carry out accurate and timely diagnosis using the standard microscopy method. Where lab technicians are
available, the results are usually subjective due to variations in expert judgement. To address this challenge, prompt
interventions to improve disease control are needed. The emerging technologies of machine learning that can learn com-
plex image patterns have accelerated research in medical image analysis. In this study, on a dataset of thick blood smear
images, we evaluate and compare performance of three pre-trained deep learning architectures namely; faster regional
convolutional neural network (faster R-CNN), single-shot multi-box detector (SSD) and RetinaNet through a Tensorow
object detection API. Data augmentation method was applied to optimise performance of the meta architectures. The
possibility for mobile phone detector deployment was also investigated. The results revealed that faster R-CNN was the
best trained model with a mean average precision of over 0.94 and SSD, was the best model for mobile deployment. We
therefore deduce that faster R-CNN is best suited for obtaining high rates of accuracy in malaria detection while SDD is
best suited for mobile deployment.
Keywords Deep learning· Malaria detection· Thick blood smear· Mobile detector
1 Introduction
Malaria is one of the top causes of death in sub-Saharan
Africa[1]. Of the 438,000 malaria cases registered, an
estimated 92% resulted in deaths; two-thirds of which
occurred among children under 5years of age[1]. In highly
endemic areas like Uganda, Malaria is the leading cause of
death accounting for over 27% of loss of lives[2]. There-
fore, detecting malaria parasites is key to malaria diagnosis
as it contributes immensely to the prevention and treat-
ment of the deadly disease[3].
The conventional approach for the diagnosis of malaria
is microscopy[3, 4] that requires interpretation of the
results by a skilled technician. Using this approach, a
sample of blood is drawn from a patient and smeared
on a glass slide stained with Giemsa. This blood sample
can be either thin or thick. A thick smear is more ecient
for parasite detection with an 11 times higher sensitivity
rate[5]. Although higher accuracy has been recorded with
thick blood smears, most of the research in automation of
malaria diagnosis has been done with thin blood smears
thus calling for more research on thick smears[6].
Furthermore, there are few skilled lab technicians in the
highly endemic areas to eectively interpret the micros-
copy diagnosis results. When they do, the interpretation
of the results is aected by the level of human variabil-
ity in observation of the parasites which leads to bias in
Supported by SIDA.
* Rose Nakasi, g.nakasirose@gmail.com | 1College ofComputing andInformation Sciences, Makerere University, Kampala,
Uganda. 2Makerere AI Research, Kampala, Uganda.
Content courtesy of Springer Nature, terms of use apply. Rights reserved.