Poster Presentation at the European Congress of Radiology, Vienna, 2019
To evaluate the performance of a deep learning system based on convolutional neural networks in characterizing pulmonary nodules and predicting the presence of malignancy on chest CT scans. We also attempt to benchmark its performance against four radiologists.
Methods and Materials
A deep-learning system based on Convolutional Neural Networks was trained on 1245 scans from the NLST trial with pathologically proven ground truths to determine malignancy status of lung cancers. 100 unseen low-dose CT scans from the validation set were chosen at random and predictions were generated from the system. Studies were randomized and presented to 4 thoracic radiologists with 2, 5, 8- and 15-years’ experience to characterize the nodules. The radiologists were asked to assess the probability of malignancy in the scans on a Likert scale of 1 (highly unlikely) to 5 (highly likely). The AUC curves were analysed for the Algorithm and the radiologists.
The radiologists had AUCs of 0.79, 0.8, 0.82 and 0.83. Individually, radiologists’ accuracy varied from 73 to 74 % and AI’s accuracy was 79 % with an AUC of 0.86. The difference in the radiologist’s interpretation was not found to be statistically significant as the one - way ANOVA revealed p-value is 0.77311
The deep learning system shows better performance than experienced radiologists in predicting the presence of malignant nodule on the cases obtained from the NLST dataset.