Fig 3. Site-level analysis.

Flow chart which demonstrates the quantitative analysis carried out on the site specific ROIs. The purpose of this analysis was to develop a predictive model to differentiate tumor tissue from muscle and adipose. A total of 30 images were used as a training dataset to develop the model. Each image was further divided into smaller bins to ensure very localized disease could be detected. Finally, area fraction, average diameter, and average shape were calculated for each bin and used as input variables for a logistic regression model. Scale bar is 200 μm.