6 research outputs found
An optimised YOLOv4 deep learning model for efficient malarial cell detection in thin blood smear images
No full textAbstract Background Malaria is a serious public health concern worldwide. Early and accurate diagnosis is essential for controlling the disease’s spread and avoiding severe health complications. Manual examination of blood smear samples by skilled technicians is a time-consuming aspect of the conventional malaria diagnosis toolbox. Malaria persists in many parts of the world, emphasising the urgent need for sophisticated and automated diagnostic instruments to expedite the identification of infected cells, thereby facilitating timely treatment and reducing the risk of disease transmission. This study aims to introduce a more lightweight and quicker model—but with improved accuracy—for diagnosing malaria using a YOLOv4 (You Only Look Once v. 4) deep learning object detector. Methods The YOLOv4 model is modified using direct layer pruning and backbone replacement. The primary objective of layer pruning is the removal and individual analysis of residual blocks within the C3, C4 and C5 (C3–C5) Res-block bodies of the backbone architecture’s C3-C5 Res-block bodies. The CSP-DarkNet53 backbone is simultaneously replaced for enhanced feature extraction with a shallower ResNet50 network. The performance metrics of the models are compared and analysed. Results The modified models outperform the original YOLOv4 model. The YOLOv4-RC3_4 model with residual blocks pruned from the C3 and C4 Res-block body achieves the highest mean accuracy precision (mAP) of 90.70%. This mAP is > 9% higher than that of the original model, saving approximately 22% of the billion floating point operations (B-FLOPS) and 23 MB in size. The findings indicate that the YOLOv4-RC3_4 model also performs better, with an increase of 9.27% in detecting the infected cells upon pruning the redundant layers from the C3 Res-block bodies of the CSP-DarkeNet53 backbone. Conclusions The results of this study highlight the use of the YOLOv4 model for detecting infected red blood cells. Pruning the residual blocks from the Res-block bodies helps to determine which Res-block bodies contribute the most and least, respectively, to the model’s performance. Our method has the potential to revolutionise malaria diagnosis and pave the way for novel deep learning-based bioinformatics solutions. Developing an effective and automated process for diagnosing malaria will considerably contribute to global efforts to combat this debilitating disease. We have shown that removing undesirable residual blocks can reduce the size of the model and its computational complexity without compromising its precision. Graphical Abstrac
An optimised YOLOv4 deep learning model for efficient malarial cell detection in thin blood smear images
Get PDFBackground
Malaria is a serious public health concern worldwide. Early and accurate diagnosis is essential for controlling the disease’s spread and avoiding severe health complications. Manual examination of blood smear samples by skilled technicians is a time-consuming aspect of the conventional malaria diagnosis toolbox. Malaria persists in many parts of the world, emphasising the urgent need for sophisticated and automated diagnostic instruments to expedite the identification of infected cells, thereby facilitating timely treatment and reducing the risk of disease transmission. This study aims to introduce a more lightweight and quicker model—but with improved accuracy—for diagnosing malaria using a YOLOv4 (You Only Look Once v. 4) deep learning object detector.
Methods
The YOLOv4 model is modified using direct layer pruning and backbone replacement. The primary objective of layer pruning is the removal and individual analysis of residual blocks within the C3, C4 and C5 (C3–C5) Res-block bodies of the backbone architecture’s C3-C5 Res-block bodies. The CSP-DarkNet53 backbone is simultaneously replaced for enhanced feature extraction with a shallower ResNet50 network. The performance metrics of the models are compared and analysed.
Results
The modified models outperform the original YOLOv4 model. The YOLOv4-RC3_4 model with residual blocks pruned from the C3 and C4 Res-block body achieves the highest mean accuracy precision (mAP) of 90.70%. This mAP is > 9% higher than that of the original model, saving approximately 22% of the billion floating point operations (B-FLOPS) and 23 MB in size. The findings indicate that the YOLOv4-RC3_4 model also performs better, with an increase of 9.27% in detecting the infected cells upon pruning the redundant layers from the C3 Res-block bodies of the CSP-DarkeNet53 backbone.
Conclusions
The results of this study highlight the use of the YOLOv4 model for detecting infected red blood cells. Pruning the residual blocks from the Res-block bodies helps to determine which Res-block bodies contribute the most and least, respectively, to the model’s performance. Our method has the potential to revolutionise malaria diagnosis and pave the way for novel deep learning-based bioinformatics solutions. Developing an effective and automated process for diagnosing malaria will considerably contribute to global efforts to combat this debilitating disease. We have shown that removing undesirable residual blocks can reduce the size of the model and its computational complexity without compromising its precision
Machine and deep learning methods in identifying malaria through microscopic blood smear : A systematic review
Get PDFDespite the persistency of World Health Organization to eliminate malaria since 1987, malaria disease continues to pose a significant threat to global health. As the severity of malaria persists over the years, there is a critical need for an automated diagnosis system for more efficient diagnosis and effective treatment administration. To mitigate the increase in mosquito-borne diseases, there has been a heightened interest in the application of Artificial intelligence (AI), specifically deep learning. With the assistance of the Preferred Reporting Items for
Systematic Review and Meta-Analyses (PRISMA) framework, an extensive review of current state of automated malaria diagnosis systems utilizing machine learning and deep learning approaches was performed across eight scientific databases, with 50 articles shortlisted from the years 2015–2023. Besides, identifying the research gaps, we synthesise the existing literature, analyse the outcomes, and explore the critical parameters that influence model performance. From the review, the prevailing models primarily focus on binary classification while disregarding cross-dataset validations and multi-stage classification. This gap challenges the delivering effective treatments, especially considering potential drug resistance. Established protocols and classification models are needed to anticipate the specific malaria species. The keywords in automated malaria diagnosis that
we identified include machine learning, deep learning, transfer learning, and convolutional neural networks. Through examinations of the constraints in current methodologies, we provide valuable suggestions that could propel the field of automated malaria diagnosis. This systematic review provides a comprehensive overview, critical insights, and a roadmap for future research endeavours in this vital domain of healthcare
Automated Identification of Malaria-Infected Cells and Classification of Human Malaria Parasites Using a Two-Stage Deep Learning Technique
No full textThe gold standard for diagnosing malaria remains microscopic examination; however, its application is frequently impeded by the lack of a standardized framework that guarantees uniformity and quality, particularly in scenarios with limited resources and high volume. This study suggests a novel and highly effective automated diagnostic approach that employs deep-learning object detectors to improve the accuracy and efficiency of malaria-infected cell detection and Plasmodium species classification to overcome these challenges. Plasmodium parasites were detected within thin blood stain images using the YOLOv4 and YOLOv5 models, which were optimized for this purpose. YOLOv5 obtains a slightly higher accuracy on the source dataset (mAP@ %) than YOLOv4 (mAP@ %), but YOLOv4 exhibits superior robustness and generalization across diverse datasets, as demonstrated by its performance on an independent validation set (mAP@ %). This robustness emphasizes the dependability of YOLOv4 for deployment in a variety of clinical settings. Furthermore, an automated process was implemented to produce bound single-cell images from YOLOv4’s localization outputs, thereby eradicating the necessity for conventional and time-consuming segmentation methods. The DenseNet-121 model, which was optimized for species identification, obtained an impressive overall accuracy of 95.5% in the subsequent classification stage, indicating excellent generalization across all malaria species. Accurate classification of Plasmodium species on microscopically thin blood films is essential for guiding appropriate therapy and preventing unnecessary anti-malarial treatments, which can lead to adverse effects and contribute to drug resistance. This research contributes to the field of automated malaria diagnosis by offering a comprehensive framework that substantially improves clinical decision-making, particularly in resource-limited environments
Automated Identification of Malaria-Infected Cells and Classification of Human Malaria Parasites Using a Two-Stage Deep Learning Technique
Get PDFThe gold standard for diagnosing malaria remains microscopic examination; however, its application is frequently impeded by the lack of a standardized framework that guarantees uniformity and quality, particularly in scenarios with limited resources and high volume. This study suggests a novel and highly effective automated diagnostic approach that employs deep-learning object detectors to improve the accuracy and efficiency of malaria-infected cell detection and Plasmodium species classification to overcome these challenges. Plasmodium parasites were detected within thin blood stain images using the YOLOv4 and YOLOv5 models, which were optimized for this purpose. YOLOv5 obtains a slightly higher accuracy on the source dataset (mAP@ 0.5=96 %) than YOLOv4 (mAP@ 0.5=89 %), but YOLOv4 exhibits superior robustness and generalization across diverse datasets, as demonstrated by its performance on an independent validation set (mAP@ 0.5=90 %). This robustness emphasizes the dependability of YOLOv4 for deployment in a variety of clinical settings. Furthermore, an automated process was implemented to produce bound single-cell images from YOLOv4’s localization outputs, thereby eradicating the necessity for conventional and time-consuming segmentation methods. The DenseNet-121 model, which was optimized for species identification, obtained an impressive overall accuracy of 95.5% in the subsequent classification stage, indicating excellent generalization across all malaria species. Accurate classification of Plasmodium species on microscopically thin blood films is essential for guiding appropriate therapy and preventing unnecessary anti-malarial treatments, which can lead to adverse effects and contribute to drug resistance. This research contributes to the field of automated malaria diagnosis by offering a comprehensive framework that substantially improves clinical decision-making, particularly in resource-limited environments
Proceedings of International Technical Postgraduate Conference 2022
Get PDFThis conference proceedings contains articles on the various research ideas of the academic & research communities presented at the International Technical Postgraduate Conference 2022 (TECH POST 2022) that was held at Universiti Malaya, Kuala Lumpur, Malaysia on 24-25 September 2022. TECH POST 2022 was organized by the Faculty of Engineering, Universiti Malaya. The theme of the conference is “Embracing Innovative Engineering Technologies Towards a Sustainable Future”. TECH POST 2022 conference is intended to foster the dissemination of state-of-the-art research from five main disciplines of Engineering: Electrical Engineering, Biomedical Engineering, Civil Engineering, Mechanical Engineering, and Chemical Engineering. The objectives of TECH POST 2022 are to bring together innovative researchers from all engineering disciplines to a common forum, promote R&D activities in Engineering, and promote the dissemination of scientific knowledge and research know-how between researchers, engineers, and students.
Conference Title: International Technical Postgraduate Conference 2022Conference Acronym: TECH POST 2022Conference Date: 24-25 September 2022Conference Location: Faculty of Engineering, Universiti Malaya, Kuala Lumpur Malaysia (Hybrid Mode)Conference Organizers: Faculty of Engineering, Universiti Malaya, Kuala Lumpur, Malaysia
