Skip to main content
Fig. 4 | Diagnostic Pathology

Fig. 4

From: Algorithm-assisted diagnosis of Hirschsprung’s disease – evaluation of robustness and comparative image analysis on data from various labs and slide scanners

Fig. 4

A diagram representing the structure of the DSS and the DSA, as well as the relationship between them. The process begins with digitally scanned whole tissue slides (a). The DSA searches through the scanned image and locates "areas of interest" which contain ganglion cell candidates (b). The DSA extracts images of each ganglion cell candidate and its immediate surroundings and also provides a score between 0 and 1 for each image, representing the level of "confidence" that the candidate is indeed a ganglion cell. The images containing the candidates with the highest scores are presented to the pathologist in up to 12 sets of 3 images each (d). The pathologist provides scores between 1 (no ganglion cells) and 5 (definite ganglion cells) to each image set. The function of the DSA is now complete. The DSS is in fact the combination of the pathologist score along with the DSA score. The DSS classifies each case according to a set of decision criteria as positive, negative or in doubt as follows: 1. Positive (non-HSCR)— the pathologist gave a score of 5 to any two (or more) image sets 2. Negative (HSCR)—The criterion for "Positive" is not met AND the average AI score is < 0.6. 3. In doubt (requires expert consultation)—The criterion for "Positive" is not met AND the average AI score is ≥ 0.6

Back to article page