From: High-throughput quantitative histology in systemic sclerosis skin disease using computer vision
Term | Analysis tool | Purpose |
---|---|---|
Image Patch Score | Principal Component Analysis was applied to the 4096 Quantitative Image Features (QIF) generated by the deep neural network (DNN) algorithm for each of the 100 image patches/biopsy | To quantitatively summarize the variance in SSc biopsy histology |
Biopsy Score | The mean of the 100 Image Patch Scores for each biopsy section | Used as a discovery tool to assess the utility of applying DNN algorithms to stained SSc biopsies. Defining the Biopsy score as the mean of the 100 Image Patch Scores enabled generation of one quantitative histologic score for each biopsy section |
Diagnostic Score | Logistic regression | To identify QIF that are associated with SSc versus health control biopsy |
Fibrosis Score | Linear regression | To identify QIF that are associated with mRSS |