Fig. 1 | Scientific Reports

Fig. 1

From: A whole slide image-based risk score predicts prognosis and postmastectomy radiotherapy benefit in triple negative breast cancer patients

Fig. 1

The patient selection process and schematic diagram of experimental design. (a) The flow chart of the patient selection process. (b) The schematic diagram of experimental design (created by BioRender.com). WSIs were divided into small tiles for energy analysis firstly. High-energy tiles were retained for further tumor cell segmentation and cell features quantification. Thus, the image score was calculated by a XGBoost model using important features selected by Cox regression analysis. Similar methods were applied to the clinicopathological score (containing age, T stage, N stage) and the risk score (containing the image score and age, T stage, N stage). Finally, patients were divided into “high-risk” and “low-risk” groups according to the risk score, and these patients could or could not benefit from PMRT. (c) The top 80 high energy value tiles (top) and the low energy value tiles (bottom) calculated using Sobel filter. Scale bar: 100 μm. Mask-RCNN, the mask region convolutional neural network; PMRT postmastectomy radiotherapy.

Back to article page