Abstract
Introduction/Objective
Objective: This work investigates utilisation of deep learning enabled peripheral blood smear image analysis for automated detection and enumeration of red cell parasites.
Methods/Case Report
Methods: Peripheral blood smear red cell images from 30 individuals identified as positive and/or negative for plasmodilum falciparum forms were used. Blood submitted to hematology laboratory for complete blood count was evaluated on Sysmex XN 3100 analyzer with routine peripheral blood slides performed on instrument associated SP50 stainer. Images of red cells obtained with Xfinity DX40 microscope mounted camera were subjected to classification using deep learning software (Cognex, ViDi Suite 4.1). The training of the classification tool was performed on 200 peripheral blood smear images divided in two dataset classes: normal/negative and abnormal/positive for plasmodia, 50% of training images represented positive data set. Performance of the developed model was tested on 300 images including 66% positive for plasmodia obtained from 20 patients. Enumeration of parasitic forms was performed for each case. Model performance was compared to expert hematopathology reviewer which was used as gold standard.
Results (if a Case Study enter NA)
Results: Overall, Cognex ViDi Suite 4.1 demonstrates the effectiveness in discriminating between images positive and negative for red cell plasmodial forms as well as enables parasite quantification. Following performance specifications were determined for parasite detection: sensitivity (0.969230769), specificity (0.99383217). High correlation coefficients (0.9961) between automatically detected parasites and ground truth, on both image level and patient level, demonstrate the practicality of our method.
Conclusion
Deep learning enabled image analysis of peripheral blood smears is a promising alternative to manual identification and enumeration of red cell plasmodial forms with performance comparable to expert hematopathology reviewer.
Peripheral blood smear image analysis: A comprehensive review