AUTHOR=Lin Qian , Wu Hai Jun , Song Qi Shi , Tang Yu Kai 

TITLE=CT-based radiomics in predicting pathological response in non-small cell lung cancer patients receiving neoadjuvant immunotherapy

JOURNAL=Frontiers in Oncology

VOLUME=Volume 12 - 2022

YEAR=2022

URL=https://www.frontiersin.org/journals/oncology/articles/10.3389/fonc.2022.937277

DOI=10.3389/fonc.2022.937277

ISSN=2234-943X

ABSTRACT=Objectives: In radiomics, high-throughput algorithms extract objective quantitative features from medical images. In this study, we evaluated CT-based radiomics features, clinical features, deep learning features, and a combination of features for predicting good pathological response (GPR) in non-small cell lung cancer (NSCLC) patients receiving immunotherapy-based neoadjuvant therapy（NAT）.
Methods and materials: We reviewed 62 patients with NSCLC who received surgery after NAT and collected clinicopathological data and CT images before and after immunotherapy-based NAT. A series of image preprocessing was carried out on CT images: tumor segmentation, radiomics feature extraction, deep learning feature extraction, and normalization. Spearman correlation coefficient (Spearman), Principal Component Analysis (PCA), Least absolute shrinkage, and selection operator (Lasso) were used to screen features. The pre-treatment traditional radiomics combined with clinical characteristics (before_rad_cil) model and pre-treatment deep learning characteristics (before_dl) model were constructed according to the data collected before treatment. The data collected after neoadjuvant therapy created the after_rad_cil model and after_dl model. The entire model was jointly constructed by all clinical features, conventional radiomics features, and deep learning features before and after neoadjuvant treatment. Finally, according to the data obtained before and after treatment, the before_nomogram and after_nomogram were constructed.
Results: 
In before_rad_cil model, six features were screened out to predict good pathological response (GPR). The average prediction accuracy (ACC) after modeling with k-Nearest Neighbor (KNN) was 0.707. 
In after_rad_cil model, nine features predictive of GPR were obtained after feature screening; The ACC after modeling with Support Vector Machine（SVM） was 0.682. 
Before_dl model and after_dl model were modeled by SVM, and the ACC was 0.629 and 0.603, respectively. 
After feature screening, the entire model was constructed by Multilayer Perceptron (MLP), and the average prediction accuracy of GPR was the highest, 0.805. 
The Calibration curve showed that the predictions of GPR by before_nomogram and after_nomogram were in consensus with the actual GPR.
Conclusion: 
CT-based radiomics has a good predictive ability for good pathological response in non-small cell lung cancer patients receiving immunotherapy-based neoadjuvant. Among the radiomics features combined with the clinicopathological information model, deep learning features model, and the entire model, the entire model had the highest prediction accuracy.