Table 4.

Comparison of classification accuracy using either BAF or OCT images for each of the 4 described methods.

	BAF images	OCT images	p
UP hb	0.614 (CI 0.557–0.672)	0.662 (CI 0.657–0.684)	0.031
P hb	0.785 (CI 0.768–0.810)	0.741 (CI 0.718–0.757)	0.025
UP CNN	0.861 (CI 0.843–0.879)	0.867 (CI 0.853–0.881)	0.652
P CNN	0.880 (CI 0.862–0.896)	0.893 (CI 0.882–0.911)	0.790

The analysis responds to the question “Does UPhb/Phb/UP CNN/P CNN perform better when applied on BAF or OCT images?”. BAF blue autofluorescence, CNN convolutional neural network, OCT optical coherence tomography, P CNN CNN classification of unprocessed images, P hb human-based classification of processed images, UP CNN CNN classification of unprocessed images, UP hb human-based classification of unprocessed images.