TABLE 2.

Classification Results for Individual Patches. The Validation Cohort of Images (N = 63 Active EoE; N = 63 non-EoE) Was Subjected to Cropping Into Patches With the Indicated Pixel Sizes and Downscaled When Indicated. True Positive Rate (TPR; Number of Patches Classified as Active EoE / Number of Active EoE Patches X 100), True Negative Rate (TNR; Number of Patches Classified as non-EoE / Number of non-EoE Patches X 100), Accuracy (Number of Patches Accurately Classified as Either Active EoE or non-EoE / Total Number of Patches X 100), and Predicted Prevalence (Total Number of Images Classified as Active [i.e., True Positive + False Positive Number of Images] / Total Number of Images) for Each Patch Size and Downscaling Method (If Applicable) are Shown. DCNN, Deep Convolutional Neural Network; TPR, True Positive Rate; TNR, True Negative Rate. ACC, Accuracy

Original Image	Final DCNN input image size	Active EoE (TPR)	Non-EoE (TNR)	ACC	Predicted Prevalence (PP)
Patch = 448×448	224×224 (Downscale)	77.0%	79.7%	78.3%	0.49
Patch = 224×224	224×224	73.3%	75.2%	74.2%	0.49