Fig. 3

The performance of different models. A, B ROC curves of different models for predicting pathological complete response (pCR) of adenocarcinoma of the esophagogastric junction (AEG) in the training group (A) and external validation group (B). C, D Calibration curves of different models predicting pCR in the training group (C) and external validation group (D). The 45-degree sloping line indicates the ideal calibration, and the closer the model calibration curve is to the ideal calibration line, the better the agreement between the model predicted probability and the actual probability. E, F Decision curves of different prediction models in the training group (E) and external validation group (F). The X-axis is the threshold probability range, and the Y-axis is the net benefit. The black line labeled "NONE" indicates that no lesions are assumed to be pCR, and the gray line labeled "ALL" indicates that all lesions are assumed to be pCR. The further away from both the black and gray lines, the higher the net benefit of the model compared to performance utilizing the "NONE" and "ALL" assumptions. When comparing the decision curves of different models within the same range of threshold probability, the larger the area under the curve for the same threshold probability interval, the higher the net benefit of the model at that threshold probability