Team: Gordon Center for Medical Imaging, Massachusetts General Hospital, Harvard Medical School, USA
Authors: Aoxiao Zhong, Quanzheng Li
I have reproduced the algorithm of the top-performing team in the challenge event (HMS-MIT Method 1). However, instead of training the GoogLeNet from scratch, I used the pre-trained GoogLeNet model trained for ImageNet classification to initialize the weights and then finetuned the network. Similar strategies were taken to produce scores at the Image and lesion levels.
The following figure shows the receiver operating characteristic (ROC) curve of the method.
The following figure shows the free-response receiver operating characteristic (FROC) curve of the method.
The table below presents the average sensitivity of the developed system at 6 predefined false positive rates: 1/4, 1/2, 1, 2, 4, and 8 FPs per whole slide image.
Error rendering graph from file