放射線科医の読影戦略を取り入れたAIが肺結節分類を改善（AI Model Improves Pulmonary Nodule Classification by Following Radiologists’ Reading Strategy）

2026-09-29

2026-09-28 合肥物質科学研究院（HFIPS）

中国科学院合肥物质科学研究院の研究チームは、胸部CT画像から肺結節の良悪性を分類する3次元AIモデル「M³Net」を開発した。放射線科医が「肺全体→結節周辺→結節内部」と段階的に画像を読む方法を模倣し、異なるスケールの情報を統合する点が特徴である。公開データセットLIDC-IDRIと中国科学技術大学の臨床データセットで評価した結果、分類精度はそれぞれ86.96％、84.24％となり、比較した12手法の中で最高性能を示した。また、Grad-CAMによる可視化で、モデルが診断に関係する領域に注目していることも確認された。肺がんの早期発見における肺結節評価を支援する、説明可能性を備えたコンピューター支援診断技術への応用が期待される。

Workflow of the clinically inspired M³Net model for pulmonary nodule classification (Image by GE Dianlong)

＜関連情報＞

M3Net ：肺結節分類のためのマクロ→メソ→ミクロの臨床的着想に基づく階層型3Dネットワーク M3Net: A macro → meso → micro clinical-inspired hierarchical 3D network for pulmonary nodule classification

Jinyue Li, Yuzhou Yu, Jingjing Yang, Meng Fu, Yani Zhang, Shuyao He, Dianlong Ge, Xin Ning, Yannan Chu, Qiankun Li
Information Fusion Available online: 29 March 2026
DOI:https://doi.org/10.1016/j.inffus.2026.104334

Highlights

A clinically guided Macro-Meso-Micro (M³) 3D network enhances explainability in nodule classification.
Progressive multi-scale fusion aligns model reasoning with radiologists’ hierarchical logic.
Hierarchical cross-attention enables progressive reasoning across local and global cues.
Experiments on LIDC-IDRI and clinical data show state-of-the-art performance and robustness.

Abstract

The accurate classification of benign and malignant pulmonary nodules in CT scans is critical for early lung cancer screening, yet remains challenging due to the multi-scale and heterogeneous nature of pulmonary nodules. While deep learning offers potential for auxiliary diagnosis, most existing models act as ”black boxes”, lacking the transparency and explainability required for trustworthy clinical integration. To address this issue, we propose M³Net, a novel 3D network for pulmonary nodule classification inspired by the hierarchical diagnostic workflow of radiologists, which integrates multi-scale contextual information from fine-grained structures to global anatomical relationships. Our framework constructs a progressive multi-scale input, from fine-grained nodule structures to local semantics and global spatial relationships. M³Net employs scale-specific encoders and ensures cross-scale semantic consistency through latent space projection and mutual information maximization. Extensive experiments on the public LIDC-IDRI dataset and a self-collected clinical dataset (USTC-FHLN) demonstrate that our method achieves state-of-the-art performance, with accuracies of 86.96% and 84.24% respectively, outperforming the best baseline by 3.26% and 2.17%. The results validate that M³Net provides a more robust and clinically relevant solution for pulmonary nodule classification. The code is available at https://github.com/jylEcho/M3-Net.