Fig. 7: Biases of tumor-infiltrating lymphocyte (TIL) evaluation by pathologists and deep learning (DL)-powered TIL analyzer. | npj Breast Cancer

Fig. 7: Biases of tumor-infiltrating lymphocyte (TIL) evaluation by pathologists and deep learning (DL)-powered TIL analyzer.

From: Deep learning model improves tumor-infiltrating lymphocyte evaluation and therapeutic response prediction in breast cancer

Fig. 7

a Classification of the initial discordance cases of TIL interpretation by pathologists. b Degree of achieving concordance by the DL assistance according to the initial discordance categories. c Example of an uneven stromal TIL (sTIL) distribution. The area in the blue box represents a low sTIL area, while the area in the red box represents a high sTIL area (scale bar, 250 μm). d Classification of the estimation error by DL-powered TIL analyzer. e Classification of the DL misinterpretations by causes. f Example of missed lymphoid cells by the DL-powered TIL analyzer (scale bar, 100 μm). g Example of incorrect cancer stroma segmentation by the DL-powered TIL analyzer (scale bar, 250 μm). Red dot, lymphocytes detected by DL model; white arrowhead, lymphoid cells missed by DL model; yellow dot: tumor cells detected by DL model; green region, cancer stroma segmented by DL model; purple region, cancer area segmented by DL model.

Back to article page