Fig. 5
From: Interpretable model based on MRI radiomics to predict the expression of Ki-67 in breast cancer
Visualization of the model’s predictions by SHAP, where the base value signifies the model’s average predicted value without any feature influence. Arrow: contribution of each feature, pointing in the direction of the expected outcome; length indicates the importance of the contribution. f(x): the actual prediction result of the model for a specific case. (A,B) A 51 year old patient, invasive carcinoma of the right breast, no special type, with Ki-67 of 30% and high expression predicted by the Combined model; (C,D) A 55 year old patient with invasive carcinoma of the left breast, no special type, with 10% Ki-67 and low expression as predicted by the Combined model. (B) and (D) show waterfall plots, and the dashed line in the figure shows this patient’s overall decision-making process in the model.
