Figure 1 | Scientific Reports

Figure 1

From: Deep learning-based classification of blue light cystoscopy imaging during transurethral resection of bladder tumors

Figure 1

The identification results of the malignant tumor lesions in bladder. (A) The classification sensitivities of benign and malignant images based on the considered fine-tuned CNNs and the physician ratings. All CNNs could predict the malignant images quit well with sensitivity of at least 91% and specificity larger than 77%. (B) Comparison between the mean sensitivities of the fine-tuned CNNs and the physician ratings. The MobileNetV2 network followed by VGG16 network showed the best classification results with a mean sensitivity of 91.81% and 90.75%; respectively. (C) The class distribution of the BL image data set with respect to the percentage of malignant and benign images in the data set. Clearly, the number of malignant images is much larger than the number of the images collected from benign legions.

Back to article page