Label-free imaging and classification of live P. falciparum enables high performance parasitemia quantification without fixation or staining
Fig 7
a) The Faster-RCNN method detects objects directly on images without using an intermediate semantic segmentation step [45], generating bounding boxes and confidence scores for each detected object. The image shown was acquired at 285 nm, and includes examples of healthy cells, ring stage parasites, trophozoites, as well as an echinocyte (spiky RBC, lower middle-left). b) A four-category confusion matrix is shown for the Faster R-CNN method, using the same format as in Fig 2. Faster R-CNN exhibited a reduced FPR and higher overall accuracy than the two-step method.