Development and external validation of the multichannel deep learning model based on unenhanced CT for differentiating fat-poor angiomyolipoma from renal cell carcinoma: a two-center retrospective study

Figures

• 

  Fig. 1. Flow chart of patient recruitment for the First Affiliated Hospital of Sun Yat-Sen University cohort (a) and Sun Yat-Sen University Cancer Center cohort (b), and the design for model development and validation (c)

• 

  Fig. 2. Overview of whole-tumor CT images processed by the deep learning model

• 

  Fig. 3. The performance of models trained on different CT modalities in the fivefold cross-validation. Scatter plots of the model's the image-level (a) and patient-level (b) performance in each fold. The paired T test of the mean values of image-level ACC (c) and patient-level AUC (d)

• 

  Fig. 4. Comparison of performance of the “unenhanced CT and 7-channel” model in independent internal and external validation. a The receiver operating characteristic (ROC) curves for patient-level performance on the hold-out dataset and Sun Yat-Sen University Cancer Center (SYSUCC) dataset. b The confusion matrix for the independent internal validation and external validation

• 

  Fig. 5. The t-SNE visualization was performed with image-level samples randomly selected from the hold-out dataset (a) and SYSUCC dataset (b). The red dots representing fat-poor angiomyolipoma (fp-AML) are mainly concentrated in the lower left of the coordinate system, and the blue dots representing renal cell carcinoma (RCC) are in a cord-like distribution on the right side

• 

  Fig. 6. Representative example predictions from Sun Yat-Sen University Cancer Center (SYSUCC). a An unenhanced CT image from a 50-year-old woman who was preoperatively diagnosed with renal carcinoma but eventually confirmed by pathology as fat-poor angiomyolipoma (fp-AML). Due to a misdiagnosis of the tumor, she underwent a radical nephrectomy and lost the chance to preserve her right kidney. Our model successfully identified this tumor as fp-AML. b An unenhanced CT image from a 54-year-old woman who underwent a partial nephrectomy and was pathologically confirmed as clear cell renal cell carcinoma (ccRCC). Our model successfully identified this tumor as renal cell carcinoma (RCC). ROI region of interest, CAM class activation mapping