Multi-channel multi-task deep learning for predicting EGFR and KRAS mutations of non-small cell lung cancer on CT images

Abstract

Background: Predicting the mutation statuses of 2 essential pathogenic genes [epidermal growth factor receptor (EGFR) and Kirsten rat sarcoma (KRAS)] in non-small cell lung cancer (NSCLC) based on CT is valuable for targeted therapy because it is a non-invasive and less costly method. Although deep learning technology has realized substantial computer vision achievements, CT imaging being used to predict gene mutations remains challenging due to small dataset limitations. Methods:We propose a multi-channel and multi-task deep learning (MMDL) model for the simultaneous prediction of EGFR and KRAS mutation statuses based on CT images. First, we decomposed each 3D lung nodule into 9 views. Then, we used the pre-trained inception-attention-resnet model for each view to learn the features of the nodules. By combining 9 inception-attention-resnet models to predict the types of gene mutations in lung nodules, the models were adaptively weighted, and the proposed MMDL model could be trained end-to-end. The MMDL model utilized multiple channels to characterize the nodule more comprehensively and integrate patient personal information into our learning process. Results:We trained the proposed MMDL model using a dataset of 363 patients collected by our partner hospital and conducted a multi-center validation on 162 patients in The Cancer Imaging Archive (TCIA) public dataset. The accuracies for the prediction of EGFR and KRAS mutations were, respectively, 79.43% and 72.25% in the training dataset and 75.06% and 69.64% in the validation dataset. Conclusions:The experimental results demonstrated that the proposed MMDL model outperformed the latest methods in predicting EGFR and KRAS mutations in NSCLC.

Machine Learning-Based Radiomics Signatures for EGFR and KRAS Mutations Prediction in Non-Small-Cell Lung Cancer

“…Third, our study could apply the state-of-the-art deep learning algorithm to implement the classification task. Many studies have successfully developed models to solve this problem [30,[33][34][35][36] with satisfactory results. These findings encourage us to conduct future studies using neural networks to build the baseline model.…”

Artificial Intelligence-Assisted Score Analysis for Predicting the Expression of the Immunotherapy Biomarker PD-L1 in Lung Cancer

“…DL methods have shown great advantages in image identification and classification (10), especially in cell classification (11), cancer detection (12), pathological diagnosis (13), and characterization of the spatial organization of immune cells in the TME (14). Previous studies have shown the application of DL in the analysis of multiple biomarkers in immunohistochemistry (IHC) staining, including epidermal growth factor receptor, human epidermal growth factor receptor 2, and Ki67 (15)(16)(17). AI-based quantitative diagnosis could also reduce the disadvantages of traditional methods, such as time consumption, lack of reproducibility, and interobserver variability (18,19).…”

Deep Radiotranscriptomics of Non-Small Cell Lung Carcinoma for Assessing Molecular and Histology Subtypes with a Data-Driven Analysis

“…Despite the limited number of patients for machine learning analysis, especially regarding the molecular subtype patient cohorts, the proposed radiotranscriptomic model for KRAS differentiation outperformed the model of Rizzo et al [ 23 ]. Additionally, the proposed ML-based analysis for molecular and histological subtypes outperformed the corresponding NSCLC state-of-the-art research [ 17 , 19 , 20 , 21 , 22 , 23 , 25 , 26 ] by a wide margin. The performance of the proposed radiotranscriptomics and the state-of-the-art literature is presented in detail in Table 2 .…”

“…NSCLC radiogenomic/radiotranscriptomic analyses in the current literature mainly focus on predicting molecular and histological subtypes, solely from imaging data, and correlating genomic signatures with radiomic features [ 17 , 18 , 19 , 20 , 21 , 22 , 23 , 24 , 25 , 26 ]. Only a handful of studies have combined selected radiomic and transcriptomic features into a unified predictive signature.…”