In May 2025, the Journal for ImmunoTherapy of Cancer published a pioneering study entitled “NeoPred: dual-phase CT AI forecasts pathologic response to neoadjuvant chemo-immunotherapy in NSCLC”, led by Professor Jianxing He’s team from the First Affiliated Hospital of Guangzhou Medical University / National Center for Respiratory Medicine.
The study introduces NeoPred, a multimodal artificial intelligence model that combines dual-phase CT scans (pre-treatment and pre-surgery) and clinical features to predict major pathological response (MPR) before surgery in patients undergoing neoadjuvant chemo-immunotherapy for non-small cell lung cancer (NSCLC).
【Background】
Neoadjuvant chemo-immunotherapy has become a recommended treatment for resectable or locally advanced NSCLC. However, MPR—the key indicator of treatment success—is typically assessed postoperatively, leading to treatment delays and potentially prolonged exposure to ineffective therapies.
NeoPred seeks to move the pathological endpoint forward to the preoperative phase using quantitative features from two routine CT scans and clinical data, providing thoracic surgeons with a non-invasive, real-time, and evidence-based decision tool.
【Study Design】
A multi-center study was conducted, including 509 NSCLC patients who received neoadjuvant chemo-immunotherapy from four thoracic oncology centers: 459 retrospective cases were used for model development and internal validation; 50 prospective cases were used for real-world testing.
Key steps included: Development of two separate 3D convolutional neural networks for pre-treatment and pre-surgery CT scans. Fusion of imaging and clinical data (e.g., age, sex, BMI, tumor stage) into the final dual-phase model. External validation using 59 independent cases from collaborating institutions. Human-AI comparison: Nine thoracic surgeons evaluated CT scans before and after reviewing the AI prediction maps to assess human-AI collaboration benefit.
【Key Findings】
1. Robustness:
In the external test set (n=59), NeoPred achieved AUC = 0.772 with imaging only.
When clinical features were included, the AUC increased to 0.787.
2. Real-World Performance:
In the prospective test set (n=50), NeoPred outperformed surgeons (AUC 0.760 vs. 0.720).
With AI support, surgeons improved to AUC = 0.829, and diagnostic accuracy rose to 82%.
3. Blind Spot Complementarity:
Among patients classified as “Stable Disease” (SD) by RECIST, NeoPred still identified MPR cases with AUCs of 0.742 (external) and 0.833 (prospective).
This suggests NeoPred can detect “pseudo-stable” responders using morphological and dynamic features.
【Clinical Significance】
NeoPred enables:
Earlier evidence-based decision-making, allowing prediction of MPR 1–2 weeks before surgery.
Improved accuracy in complex cases, providing a reliable second opinion.
Faster and more efficient multidisciplinary team (MDT) workflows through quantitative AI-generated metrics.
Together, these advantages may significantly enhance perioperative management and personalized treatment planning in NSCLC.
【The AI Ecosystem Built by Prof. He’s Team】
NeoPred is part of a comprehensive AI toolbox developed by Prof. Jianxing He’s team, covering the entire lung cancer management continuum:
1. Early Detection & Risk Stratification:
PulmoSeek Plus3: Combines cfDNA methylation and low-dose CT for early screening, effectively reducing overtreatment.
LcProt7: Uses plasma proteomics for non-invasive screening, staging, and recurrence prediction.
2. Intelligent Diagnosis & Histologic Classification:
LungDiag5: Applies NLP to electronic health records (both structured and unstructured) for automatic respiratory disease diagnosis.
LungPath6: Analyzes HE slides to recognize subtypes of lung adenocarcinoma, enhancing pathologic consistency.
3. Genomic Inference & Targeted Therapy:
DeepGEM4: Predicts gene mutations from HE slides in real time, without the need for manual annotations.
Radiomics EGFR10: Uses CT-based radiomics to predict EGFR mutation status preoperatively, particularly in GGO lesions.
4. Therapeutic Response & Surgical Planning:
NeoPred1: Predicts MPR post-neoadjuvant therapy, bridging systemic therapy and surgery.
AI-guided Navigation8: Assists preoperative and intraoperative localization of deep lung nodules using robotic navigation and AI.
5. Cross-Institutional Collaboration & Data Security:
CAIMEN9: Enables federated learning across 40+ centers to improve diagnostic accuracy without sharing raw patient data.d
【References】
[1] Zheng J, Yan Z, Wang R, et al. NeoPred: dual-phase CT AI forecasts pathologic response to neoadjuvant chemo-immunotherapy in NSCLC. J Immunother Cancer. 2025;13(5):e011773. doi:10.1136/jitc-2025-011773.
[2] He J, Baxter SL, Xu J, et al. The practical implementation of artificial intelligence technologies in medicine. Nat Med. 2019;25(1):30–36. doi:10.1038/s41591-018-0307-0.
[3] He J, Wang B, Tao J, et al. Accurate classification of pulmonary nodules by a combined model of clinical, imaging, and cell-free DNA methylation biomarkers. Lancet Digit Health. 2023;5(10):e647-e656. doi:10.1016/S2589-7500(23)00125-5.
[4] Zhao Y, Xiong S, Ren Q, et al. Deep learning using histological images for gene mutation prediction in lung cancer: a multicentre retrospective study. Lancet Oncol. 2025;26(1):136–146. doi:10.1016/S1470-2045(24)00599-0.
[5] Liang H, Yang T, Liu Z, et al. LungDiag: Empowering artificial intelligence for respiratory diseases diagnosis based on electronic health records, a multicenter study. MedComm (2020). 2025;6(1):e70043. doi:10.1002/mco2.70043.
[6] Huang H, Yan Z, Li B, et al. LungPath: artificial intelligence-driven histologic pattern recognition for improved diagnosis of early-stage invasive lung adenocarcinoma. Transl Lung Cancer Res. 2024;13(8):1816–1827. doi:10.21037/tlcr-24-258.
[7] Liang H, Wang R, Cheng R, et al. LcProt: Proteomics-based identification of plasma biomarkers for lung cancer multievent, a multicentre study. Clin Transl Med. 2025;15(1):e70160. doi:10.1002/ctm2.70160.
[8] Liu J, Jiang Y, He R, et al. Robotic-assisted navigation system for preoperative lung nodule localization: a pilot study. Transl Lung Cancer Res. 2023;12(11):2283–2293. doi:10.21037/tlcr-23-493.
[9] Tang R, Liang H, Guo Y, et al. Pan-mediastinal neoplasm diagnosis via nationwide federated learning: a multicentre cohort study. Lancet Digit Health. 2023;5(9):e560–e570. doi:10.1016/S2589-7500(23)00106-1.
[10] Cheng B, Deng H, Zhao Y, et al. Predicting EGFR mutation status in lung adenocarcinoma presenting as ground-glass opacity: utilizing radiomics model in clinical translation. Eur Radiol. 2022;32(9):5869–5879. doi:10.1007/s00330-022-08673-y.
END