Multi-contrast machine learning improves schistosomiasis diagnostic performance

Abstract Schistosomiasis currently affects over 250 million people and remains a public health burden despite ongoing global control efforts. Conventional microscopy is a practical tool for diagnosis and screening of Schistosoma haematobium , but identification of eggs requires a skilled microscopist. Here we present a machine learning (ML)-based strategy for automated detection of S. haematobium that combines two imaging contrasts, brightfield (BF) and darkfield (DF), to improve diagnostic performance. We collected BF and DF images of S. haematobium eggs in patient samples from two different field studies in Côte d’Ivoire using a mobile phone-based microscope, the SchistoScope. We then trained separate egg-detection ML models and compared the patient-level performance of BF and DF models alone to Boolean combinations of BF and DF models, using annotations from trained microscopists as the gold standard. We found that models trained on DF images, and almost all BF and DF combinations, performed significantly better than models trained on BF images only. When models were trained on images from the first field study, patient-level classification performance for images from the second study met the WHO Diagnostic Target Product Profile (TPP) sensitivity and specificity for the monitoring and evaluation use case. When we used images from both field studies for the training set, performance of the models was improved. This work shows that multi-contrast imaging can increase information available for classification tasks, while retaining the portability, power, and time-to-results of the TPP’s desired diagnostic. The imaging contrasts used here require no additional sample preparation and do not increase the complexity of the imaging system, and we used off-the-shelf ML models to simplify software engineering. Multi-contrast machine learning offers a practical means to improve performance of automated diagnostics for S. haematobium , one that could be applied to other microscopy-based diagnostics. Author summary Schistosomiasis is a neglected tropical disease that impacts hundreds of millions of people worldwide. Patients with Schistosoma haematobium shed parasite eggs in their urine, which can be used as a diagnostic marker of disease. However, identification of those eggs in patient samples normally requires a microscope and trained microscopist. In this work, we show that machine learning models trained on two imaging contrasts, brightfield and darkfield, can improve performance of automated schistosomiasis diagnosis. Using a mobile phone-based microscope (the SchistoScope), we captured brightfield and darkfield images of patient samples during two visits to Côte d’Ivoire and then trained models to detect eggs in the brightfield and darkfield images. When training on images from one visit and testing on images from the other visit, we find that combining the brightfield and darkfield model outputs improved the diagnostic sensitivity and specificity compared to brightfield alone, meeting the target metrics for monitoring and evaluation of schistosomiasis control programs outlined by the World Health Organization. This use of multi-contrast machine learning with a mobile microscope has the potential to improve diagnostic testing for schistosomiasis and could be extended to other neglected tropical diseases.

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.