Lymphatic Filariasis Detection Using Image Analysis

Abstract

Elephantiasis is generally detected through microscopic examination of blood. Until now, this has been difficult because microfilariae only appear in the blood at night for a few hours (nocturnal periodicity). The lack of trained microscopy technicians is a serious problem. Due to the repetitive and tedious nature of diagnosis and the fact that there are few positive cases in a population of thousands. This is a contributing factor to increased detection errors. The main problem encountered is the high degree of difficulty and precision and the long time it takes to perform laboratory examinations. Image analysis method can be used as a way to identify Lymphatic Filariasis worms in the blood. Based on the description above, it can be said that the detection of Lymphatic Filariasis worms can be done with digital image analysis. This research will use the feature extraction method and Convolutional Neural Network to identify object features in the form of worms that cause elephantiasis (Lymphatic Filariasis) in digital images recorded by Trinocular digital microscope cameras. This study aims to determine the performance of image analysis methods used in the identification process of Lymphatic Filariasis worms using digital images recorded by a Compound Trinocular microscope.

Edge Artificial Intelligence (AI) for real-time automatic quantification of filariasis in mobile microscopy

“…Developed with 33 mice samples with 1314 parasites (80% for training and 20% for testing), achieving a precision of 87.6% and a recall of 90.5% [ 50 ]. Very few studies attempted to automate filarias parasite detection detecting microfilariae, without distinguishing species [ 22 , 34 ]. The only preliminary proof of concept study that uses AI to detect microfilariae is proposed by Elvana et.…”

“…The only preliminary proof of concept study that uses AI to detect microfilariae is proposed by Elvana et. al, who used a small database with 210 images and a custom CNN with 8 convolutional layers, achieving 70% accuracy [ 34 ].…”

“…The model trained with 15,700 images from 785 patients, detects visible components on human feces using an object detection algorithm with ResNet 152 as backbone, achieving a mean average precision , achieving a precision of 87.6% and a recall of 90.5% [50]. Very few studies attempted to automate filarias parasite detection detecting microfilariae, without distinguishing species [22,34]. The only preliminary proof concept study that uses AI to detect microfilariae is proposed by Elvana et.…”

“…They correlated the automated counts with manual counts and achieved 94% specificity and 100% sensitivity [22]. Elvana et al presented a lymphatic filariasis detection system using CNN, achieving an accuracy of 70% [34]. As far as we know, there have been very few attempts to deploy edge-AI systems using deep learning which are able to support in real-time and without connectivity the analysis of optical microscopy images for filaria detection with species differentiation and more broadly NTDs diagnostics.…”