Explainable multimodal deep learning for predicting thyroid cancer lateral lymph node metastasis using ultrasound imaging

Abstract Preoperative prediction of lateral lymph node metastasis is clinically crucial for guiding surgical strategy and prognosis assessment, yet precise prediction methods are lacking. We therefore develop Lateral Lymph Node Metastasis Network (LLNM-Net), a bidirectional-attention deep-learning m...

Full description

Saved in:
Bibliographic Details
Main Authors: Pengcheng Shen, Zheyu Yang, Jingjing Sun, Yun Wang, Cheng Qiu, Yirou Wang, Yongyong Ren, Sheng Liu, Wei Cai, Hui Lu, Siqiong Yao
Format: Article
Language:English
Published: Nature Portfolio 2025-08-01
Series:Nature Communications
Online Access:https://doi.org/10.1038/s41467-025-62042-z
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1849342892499271680
author Pengcheng Shen
Zheyu Yang
Jingjing Sun
Yun Wang
Cheng Qiu
Yirou Wang
Yongyong Ren
Sheng Liu
Wei Cai
Hui Lu
Siqiong Yao
author_facet Pengcheng Shen
Zheyu Yang
Jingjing Sun
Yun Wang
Cheng Qiu
Yirou Wang
Yongyong Ren
Sheng Liu
Wei Cai
Hui Lu
Siqiong Yao
author_sort Pengcheng Shen
collection DOAJ
description Abstract Preoperative prediction of lateral lymph node metastasis is clinically crucial for guiding surgical strategy and prognosis assessment, yet precise prediction methods are lacking. We therefore develop Lateral Lymph Node Metastasis Network (LLNM-Net), a bidirectional-attention deep-learning model that fuses multimodal data (preoperative ultrasound images, radiology reports, pathological findings, and demographics) from 29,615 patients and 9836 surgical cases across seven centers. Integrating nodule morphology and position with clinical text, LLNM-Net achieves an Area Under the Curve (AUC) of 0.944 and 84.7% accuracy in multicenter testing, outperforming human experts (64.3% accuracy) and surpassing previous models by 7.4%. Here we show tumors within 0.25 cm of the thyroid capsule carry >72% metastasis risk, with middle and upper lobes as high-risk regions. Leveraging location, shape, echogenicity, margins, demographics, and clinician inputs, LLNM-Net further attains an AUC of 0.983 for identifying high-risk patients. The model is thus a promising for tool for preoperative screening and risk stratification.
format Article
id doaj-art-1527856046fd4eebb4f3f98989a82b34
institution Kabale University
issn 2041-1723
language English
publishDate 2025-08-01
publisher Nature Portfolio
record_format Article
series Nature Communications
spelling doaj-art-1527856046fd4eebb4f3f98989a82b342025-08-20T03:43:14ZengNature PortfolioNature Communications2041-17232025-08-0116111710.1038/s41467-025-62042-zExplainable multimodal deep learning for predicting thyroid cancer lateral lymph node metastasis using ultrasound imagingPengcheng Shen0Zheyu Yang1Jingjing Sun2Yun Wang3Cheng Qiu4Yirou Wang5Yongyong Ren6Sheng Liu7Wei Cai8Hui Lu9Siqiong Yao10Department of Bioinformatics and Biostatistics, School of Life Sciences and Biotechnology, Shanghai Jiao Tong UniversityDepartment of General Surgery, Ruijin Hospital, Shanghai Jiao Tong University School of MedicineDepartment of Ultrasound, Shanghai Fourth People’s Hospital, School of Medicine, Tongji UniversityHepatobiliary Pancreatic Center, Xuzhou Central HospitalMedical college, Nantong UniversityDepartment of Endocrinology and Metabolism, Shanghai Children’s Medical Center, School of Medicine, Shanghai Jiao Tong UniversityDepartment of Bioinformatics and Biostatistics, School of Life Sciences and Biotechnology, Shanghai Jiao Tong UniversityDepartment of Thyroid and Breast Surgery, Shanghai Fourth People’s Hospital, School of Medicine, Tongji UniversityDepartment of General Surgery, Ruijin Hospital, Shanghai Jiao Tong University School of MedicineDepartment of Bioinformatics and Biostatistics, School of Life Sciences and Biotechnology, Shanghai Jiao Tong UniversityDepartment of Bioinformatics and Biostatistics, School of Life Sciences and Biotechnology, Shanghai Jiao Tong UniversityAbstract Preoperative prediction of lateral lymph node metastasis is clinically crucial for guiding surgical strategy and prognosis assessment, yet precise prediction methods are lacking. We therefore develop Lateral Lymph Node Metastasis Network (LLNM-Net), a bidirectional-attention deep-learning model that fuses multimodal data (preoperative ultrasound images, radiology reports, pathological findings, and demographics) from 29,615 patients and 9836 surgical cases across seven centers. Integrating nodule morphology and position with clinical text, LLNM-Net achieves an Area Under the Curve (AUC) of 0.944 and 84.7% accuracy in multicenter testing, outperforming human experts (64.3% accuracy) and surpassing previous models by 7.4%. Here we show tumors within 0.25 cm of the thyroid capsule carry >72% metastasis risk, with middle and upper lobes as high-risk regions. Leveraging location, shape, echogenicity, margins, demographics, and clinician inputs, LLNM-Net further attains an AUC of 0.983 for identifying high-risk patients. The model is thus a promising for tool for preoperative screening and risk stratification.https://doi.org/10.1038/s41467-025-62042-z
spellingShingle Pengcheng Shen
Zheyu Yang
Jingjing Sun
Yun Wang
Cheng Qiu
Yirou Wang
Yongyong Ren
Sheng Liu
Wei Cai
Hui Lu
Siqiong Yao
Explainable multimodal deep learning for predicting thyroid cancer lateral lymph node metastasis using ultrasound imaging
Nature Communications
title Explainable multimodal deep learning for predicting thyroid cancer lateral lymph node metastasis using ultrasound imaging
title_full Explainable multimodal deep learning for predicting thyroid cancer lateral lymph node metastasis using ultrasound imaging
title_fullStr Explainable multimodal deep learning for predicting thyroid cancer lateral lymph node metastasis using ultrasound imaging
title_full_unstemmed Explainable multimodal deep learning for predicting thyroid cancer lateral lymph node metastasis using ultrasound imaging
title_short Explainable multimodal deep learning for predicting thyroid cancer lateral lymph node metastasis using ultrasound imaging
title_sort explainable multimodal deep learning for predicting thyroid cancer lateral lymph node metastasis using ultrasound imaging
url https://doi.org/10.1038/s41467-025-62042-z
work_keys_str_mv AT pengchengshen explainablemultimodaldeeplearningforpredictingthyroidcancerlaterallymphnodemetastasisusingultrasoundimaging
AT zheyuyang explainablemultimodaldeeplearningforpredictingthyroidcancerlaterallymphnodemetastasisusingultrasoundimaging
AT jingjingsun explainablemultimodaldeeplearningforpredictingthyroidcancerlaterallymphnodemetastasisusingultrasoundimaging
AT yunwang explainablemultimodaldeeplearningforpredictingthyroidcancerlaterallymphnodemetastasisusingultrasoundimaging
AT chengqiu explainablemultimodaldeeplearningforpredictingthyroidcancerlaterallymphnodemetastasisusingultrasoundimaging
AT yirouwang explainablemultimodaldeeplearningforpredictingthyroidcancerlaterallymphnodemetastasisusingultrasoundimaging
AT yongyongren explainablemultimodaldeeplearningforpredictingthyroidcancerlaterallymphnodemetastasisusingultrasoundimaging
AT shengliu explainablemultimodaldeeplearningforpredictingthyroidcancerlaterallymphnodemetastasisusingultrasoundimaging
AT weicai explainablemultimodaldeeplearningforpredictingthyroidcancerlaterallymphnodemetastasisusingultrasoundimaging
AT huilu explainablemultimodaldeeplearningforpredictingthyroidcancerlaterallymphnodemetastasisusingultrasoundimaging
AT siqiongyao explainablemultimodaldeeplearningforpredictingthyroidcancerlaterallymphnodemetastasisusingultrasoundimaging