Hypermetabolic pulmonary lesions detection and diagnosis based on PET/CT imaging and deep learning models

Sep 24, 2026

Uploaded by Aran

Deep learning models are developed to detect and classify hypermetabolic lung lesions from PET/CT imaging into four clinically relevant categories: benign, lung cancer, pulmonary lymphoma, and metastasis. A lesion localisation model is trained using manually annotated PET/CT images, while classification uses a multi-dimensional joint network combining image patches and 2D projections. Performance is evaluated on internal and external datasets and compared with radiomics, with false-positive segmentations manually reviewed. Results show strong detection and classification accuracy and clinically acceptable false positives, supporting AI-assisted clinical decision-making.

LanguageEnglish

CategoryHealthcare

FormatPDF

Pages15

Reading Time38 min

AccessFree to download

Introduction

Methods

Lesion localisation model

Classification model

Evaluation and comparison

Results

Detection performance

Classification performance

False-positive analysis

Conclusion

Graphical Abstract

Keywords

European Journal of Nuclear Medicine and Molecular Imaging (2025) 52:3792–3806 [https://doi.org/10.1007/s00259-025-07215-0](https://doi.org/10.1007/s00259-025-07215-0) Hypermetabolic pulmonary lesions detection and diagnosis based on PET/CT imaging and deep learning models Jiajia Hu1 · Ran Cheng1 · Meilin Quan2,6 · Yao Peng1 · Zi Yang3 · Qing Zhang4 · Faquan Ji5 · Yangchun Chen3 · Biao Li1 · NingWen2,6,7 Received: 5 December 2024 / Accepted: 12 March 2025 / Published online: 4 April 2025 © The Author(s) 2025 Abstract Purpose This study aims to develop and evaluate deep learning models for the detection and classification of hypermetabolic lung lesions into four categories: benign, lung cancer, pulmonary lymphoma, and metastasis. These categories are defined by their pathological origin, clinical relevance, and therapeutic implications. Methods A lesion localisation model was first developed using manually annotated PET/CT images. For classification, a multi-dimensional joint network was employed, incorporating both image patches and two-dimensional projections. Classification performance was quantified by metrics like accuracy, and compared to that of a radiomics model. Additionally, false-positive segmentations were manually reviewed and analysed for clinical evaluation. Results The study retrospectively included 647 cases (409 males/238 females) over more than 8 years from five centres, divided into an internal dataset (426 cases from Shanghai Ruijin Hospital), an external test set I (151 cases from four other institutions), and an external test set II (70 cases from a new imaging device) . The localisation model achieved detection rates of 81.19%, 75.48%, and 77.59% on the internal, external test set I, and external test set II, respectively. The classification model outperformed the radiomics approach, with area-under-curves of 88.4%, 80.7%, and 66.6%, respectively. Most false-positive segmentations were clinically acceptable, corresponding to suspicious lesions in adjacent regions, particularly lymph nodes. Conclusion Deep learning models based on PET/CT imaging can effectively detect, segment, and classify hypermetabolic lung lesions, and identify suspicious adjacent lesions. These results highlight the potential of artificial intelligence in clinical decision-making and lung disease diagnosis. Jiajia Hu, Ran Cheng, Meilin Quan contributed equally to this work. * Yangchun Chen [fudanzhsh@outlook.com](fudanzhsh@outlook.com) * Biao Li [lb10363@rjh.com.cn](lb10363@rjh.com.cn) * Ning Wen[wn12479@rjh.com.cn](wn12479@rjh.com.cn) Jiajia Hu [hjj11592@rjh.com.cn](hjj11592@rjh.com.cn) Ran Cheng [chengran354@163.com](chengran354@163.com) Meilin Quan [qml03389@rjh.com.cn](qml03389@rjh.com.cn) 1 Department of Nuclear Medicine, Ruijin Hospital, Shanghai Jiao Tong University School of Medicine, No.197 Ruijin Second Road, Huangpu District, Shanghai, China 2 Department of Radiology, Ruijin Hospital, Shanghai Jiao Tong University School of Medicine, No.197 Ruijin Second Road, Huangpu District, Shanghai, China 3 Department of Nuclear Medicine, Shanghai Pulmonary Hospital, Tongji University School of Medicine, No.507 Zhengmin Road, Yangpu District, Shanghai, China 4 Department of Nuclear Medicine, Lu’an People’s Hospital of Anhui Province, No.21 Wanxi Road, Jin’an District, Lu’an, Anhui Province, China 5 Department of Nuclear Medicine, The Affiliated Hospital of Nanjing University of Chinese Medicine, No.155 Hanzhong Road, Qinhuai District, Nanjing, Jiangsu Province, China 6 The SJTU-Ruijin-UIH Institute for Medical Imaging Technology, Ruijin Hospital, Shanghai Jiao Tong University School of Medicine, 889 Shuangding Road, Jiading District, Shanghai, China 7 The Global Institute of Future Technology, Shanghai Jiao Tong University, 800 Dongchuan Road, Minhang District, Shanghai, China Graphical Abstract Keywords Hypermetabolic pulmonary lesions · Deep learning · Positron emission tomograph

FAQs

What four categories of hypermetabolic lung lesions does the study aim to classify?

The study classifies hypermetabolic lung lesions into benign, lung cancer, pulmonary lymphoma, and metastasis.

How is the lesion localisation model built?

It is developed using manually annotated PET/CT images to localise lesions prior to classification.

How does the deep learning classification model compare with radiomics?

The classification model outperforms the radiomics approach, reporting higher area-under-curve values across the evaluation sets.

Sep 24, 2026

Uploaded by Aran