Deep learning for predicting major pathological response to neoadjuvant chemoimmunotherapy in non-small cell lung cancer: A multicentre study

Abstract

Background: This study, based on multicentre cohorts, aims to utilize computed tomography (CT) images to construct a deep learning model for predicting major pathological response (MPR) to neoadjuvant chemoimmunotherapy in non-small cell lung cancer (NSCLC) and further explore the biological basis under its prediction.

Methods: 274 patients undergoing curative surgery after neoadjuvant chemoimmunotherapy for NSCLC at 4 centres from January 2019 to December 2021 were included and divided into a training cohort, an internal validation cohort, and an external validation cohort. ShuffleNetV2x05-based features of the primary tumour on the CT scans within the 2 weeks preceding neoadjuvant administration were employed to develop a deep learning score for distinguishing MPR and non-MPR. To reveal the underlying biological basis of the deep learning score, a genetic analysis was conducted based on 25 patients with RNA-sequencing data.

Findings: MPR was achieved in 54.0% (n = 148) patients. The area under the curve (AUC) of the deep learning score to predict MPR was 0.73 (95% confidence interval [CI]: 0.58-0.86) and 0.72 (95% CI: 0.58-0.85) in the internal validation and external validation cohorts, respectively. After integrating the clinical characteristic into the deep learning score, the combined model achieved satisfactory performance in the internal validation (AUC: 0.77, 95% CI: 0.64-0.89) and external validation cohorts (AUC: 0.75, 95% CI: 0.62-0.87). In the biological basis exploration for the deep learning score, a high deep learning score was associated with the downregulation of pathways mediating tumour proliferation and the promotion of antitumour immune cell infiltration in the microenvironment.

Interpretation: The proposed deep learning model could effectively predict MPR in NSCLC patients treated with neoadjuvant chemoimmunotherapy.

Funding: This study was supported by National Key Research and Development Program of China, China (2017YFA0205200); National Natural Science Foundation of China, China (91959126, 82022036, 91959130, 81971776, 81771924, 6202790004, 81930053, 9195910169, 62176013, 8210071009); Beijing Natural Science Foundation, China (L182061); Strategic Priority Research Program of Chinese Academy of Sciences, China (XDB38040200); Chinese Academy of Sciences, China (GJJSTD20170004, QYZDJ-SSW-JSC005); Shanghai Hospital Development Center, China (SHDC2020CR3047B); and Science and Technology Commission of Shanghai Municipality, China (21YF1438200).

Keywords: Deep learning; Major pathological response; Neoadjuvant chemoimmunotherapy; Non-small cell lung cancer.

Fig. 1

Flowchart illustrates the study design. (Ⅰ) A total of 274 patients acquired from four institutions are divided into three cohorts; (Ⅱ) The region of interest is annotated with a bounding box covering the lesion on CT images; (Ⅲ) The deep learning model is built using convolutional neural network algorithm; (IV) The predictive efficiency of the deep learning model is assessed in the internal/external validation cohorts, and the underlying predictive mechanism of the deep learning model is investigated with the visual analysis and genetic analysis.

Fig. 2

The performance of the deep learning model and clinical model for predicting the MPR to neoadjuvant chemoimmunotherapy. (a) ROC curves of the deep learning model in three cohorts; (b) The box figure shows the distribution of the deep learning score between MPR and non-MPR groups in three cohorts; (c) ROC curves of the clinical model in three cohorts; (d) Histograms shown the percentage of squamous cell carcinoma between MPR and non-MPR groups in three cohorts; (e–j) ROC curves of the deep learning model based on the clinicopathologic factors, including the pretreatment clinical T stage (e), N stage (f), TNM stage (g), Histologic subtype (h), gender (i) and age (j) in the whole population; (k) Waterfall plot for deep learning score in the whole population. (l) Line chart for delta range of deep learning score and AUCs in the whole population. AUC, area under the curve; MPR, major pathological response; ROC, receiver operating characteristic curve; SCC, squamous cell carcinoma.

Fig. 3

The performance of the combined model for predicting the MPR to neoadjuvant chemoimmunotherapy. (a) ROC curves of the combined model in three cohorts; (b) DeLong test for AUCs among the clinical model, deep learning model and combined model in the training cohort; (c) DeLong test for AUCs among the clinical model, deep learning model and combined model in the internal validation cohort; (d) DeLong test for AUCs among the clinical model, deep learning model and combined model in the external validation cohort; (e) Distribution graph for two-dimensional spatial structure and diagnostic metrics of the deep learning score in the whole population; (f) Distribution graph for two-dimensional spatial structure and diagnostic metrics of the combined score in the whole population; (g) Distribution graph for two-dimensional spatial structure and diagnostic metrics of the deep learning score and combined score for patients with different predicted outcomes between two models in the whole population. AUC, area under the curve; MPR, major pathological response; ROC, receiver operating characteristic curve.

Fig. 4

The genetic analysis for investigating the underlying biological basis of the developed deep learning model. (a) Heatmap of z-score normalized gene expressions presenting the differential expressed genes in samples categorized as low deep learning score compared with that categorized as high deep learning score; (b) Volcano diagram of gene expression profiles in samples separated by low deep learning score versus high deep learning score. The red dots represent genes upregulated in patients categorized as high score, whereas the blue dots represent genes upregulated in patients categorized as low score. The x-axis denotes the fold change (log2 scale), whereas the y-axis indicates statistical significance (−log10 format); (c) Bubble plot of the top 10 enriched pathways identified by gene enrichment analysis for the set of differential expressed genes, ranked by the odds ratio; (d) Box plot representing the estimation of the abundances of member cell types in a mixed cell population. FDR, false discovery rate. (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.)