Deep Learning: A Non-invasive Model to Predict EGFR Mutation Status in Lung Adenocarcinoma
Keywords:
model; EGFR; learning; deep; DL; tumor; Figure; different
Deep Learning: A Non-invasive Model to Predict EGFR Mutation Status in Lung Adenocarcinoma
Please click here to download the model.
EGFR-Net: This study proposed a deep learning (DL) model for EGFR mutation status prediction in lung adenocarcinoma. This DL model is designed based on the DenseNet and utilized transfer learning to initialize the first 20 convolutional layers of the model. We trained the DL model in 14926 CT images from 603 lung adenocarcinoma patients in one hospital, and validated its performance in the other independent hospital including 241 patients.
The DL model achieved encouraging predictive performance in both the primary cohort (n = 603; AUC = 0.88) and the independent validation cohort (n = 241; AUC = 0.81).
To use this model, users only need to feed the region of interest (ROI) of the tumor CT image into the model. The DL model will give the EGFR-mutant probability for the tumor and highlight the suspicious tumor region that are most related to EGFR-mutation status, which is very easy to use in application.
Figure 1. The deep learning model.
Figure 2. Quantitative results of the deep learning model.
Figure 3. The suspicious tumor areas discovered by the deep learning model
Figure 4. Deep learning feature analysis.
(a) The convolutional filters (features) of the deep learning model. (b) The response of filters on different tumours. (c) The response of different filters showed significant difference in different EGFR mutation status. (d) The deep learning feature showed obvious clusters in EGFR-mutant and EGFR-wild type tumors.