NeoPred: dual-phase CT AI forecasts pathologic response to neoadjuvant chemo-immunotherapy in NSCLC

Abstract

Background: Accurate preoperative prediction of major pathological response or pathological complete response after neoadjuvant chemo-immunotherapy remains a critical unmet need in resectable non-small-cell lung cancer (NSCLC). Conventional size-based imaging criteria offer limited reliability, while biopsy confirmation is available only post-surgery.

Methods: We retrospectively assembled 509 consecutive NSCLC cases from four Chinese thoracic-oncology centers (March 2018 to March 2023) and prospectively enrolled 50 additional patients. Three 3-dimensional convolutional neural networks (pre-treatment CT, pre-surgical CT, dual-phase CT) were developed; the best-performing dual-phase model (NeoPred) optionally integrated clinical variables. Model performance was measured by area under the receiver-operating-characteristic curve (AUC) and compared with nine board-certified radiologists.

Results: In an external validation set (n=59), NeoPred achieved an AUC of 0.772 (95% CI: 0.650 to 0.895), sensitivity 0.591, specificity 0.733, and accuracy 0.627; incorporating clinical data increased the AUC to 0.787. In a prospective cohort (n=50), NeoPred reached an AUC of 0.760 (95% CI: 0.628 to 0.891), surpassing the experts' mean AUC of 0.720 (95% CI: 0.574 to 0.865). Model assistance raised the pooled expert AUC to 0.829 (95% CI: 0.707 to 0.951) and accuracy to 0.820. Marked performance persisted within radiological stable-disease subgroups (external AUC 0.742, 95% CI: 0.468 to 1.000; prospective AUC 0.833, 95% CI: 0.497 to 1.000).

Conclusions: Combining dual-phase CT and clinical variables, NeoPred reliably and non-invasively predicts pathological response to neoadjuvant chemo-immunotherapy in NSCLC, outperforms unaided expert assessment, and significantly enhances radiologist performance. Further multinational trials are needed to confirm generalizability and support surgical decision-making.

Keywords: Adjuvant; Immunotherapy; Lung Cancer.

Figure 1

Overview of the study design and workflow for the establishment and validation of NeoPred. ICI immune checkpoint inhibitor.

Figure 2

Flowchart of inclusion and exclusion. NSCLC, non-small cell lung cancer.

Figure 3

Performance and analysis of models for external validation set. (A) ROC curve of the MobileNet-based model for external validation set. (B) ROC curve of the ResNet-18-based model for external validation set. (C) ROC curve of the One-hot encoding-based model for external validation set. (D) ROC curve of the combined model incorporating dual-phase CT scans and clinical factors for external validation set. (E) ROC curve of the model augmented with data augmentation techniques (RandRotate) for external validation set. (F) Scatter plot showing the relationship between pathological response and tumor regression ratio in CT scans. (G) Violin plot showing the distribution of deep learning scores between MPR and non-MPR groups in three models for external validation set. (H) Epoch-loss curve of the model during training. AUC, area under the curve; MPR, major pathological response; pCR, pathological complete response; ROC, receiver-operating-characteristic curve.

Figure 4

Performance and analysis of models for internal validation set. (A) ROC curve of NeoPred for internal validation set. (B) Violin plot showing the distribution of deep learning scores between MPR and non-MPR groups in two models for internal validation set. (C) ROC curve of experts’ performance without NeoPred assistance for internal validation set. (D) ROC curve of experts’ performance with NeoPred assistance for internal validation set. AUC, area under the curve; MPR, major pathological response; ROC, receiver-operating-characteristic curve.

Figure 5

Performance and analysis of models in SD cohort. (A) ROC curve of the MobileNet-based model for external validation set in SD cohort. (B) ROC curve of NeoPred for internal validation set in SD cohort. AUC, area under the curve; ROC, receiver-operating-characteristic curve; SD, stable disease.