An Interpretable Machine Learning Model With Synthetic MRI-Based Habitat Radiomics for Predicting Lymph Node Metastasis in Oral Cancer

Sep 22, 2026

Develop an interpretable radiomics model based on habitat to predict lymph node metastasis in oral cancer using Synthetic MRI (SyMRI). Conduct a retrospective study of 101 patients with pathologically confirmed lymph node status, using MRI MAGiC (SyMRI) including K-means clustering of intra-tumoral VOIs into three spatial habitats. Interpret predictions with SHAP and build a radiomics nomogram integrating independent variables. Habitat radiomics with high T1 and moderately low T2 outperforms whole-tumor radiomics, improving AUC versus clinical diagnosis in both cohorts.

LanguageEnglish

CategoryHealthcare

FormatPDF

Pages12

Reading Time30 min

AccessFree to download

Abstract

Purpose

Methods

Results

Conclusion

Introduction

Background and clinical need

SyMRI and MAGiC

Quantitative T1/T2 and prior evidence

Radiomics and interpretability

Cancer Medicine | RESEARCH ARTICLE OPEN ACCESS
An Interpretable Machine Learning Model With Synthetic MRI-Based Habitat Radiomics for Predicting Lymph Node Metastasis in Oral Cancer
Rui Wang1 | Jiliang Ren1 | Yong Zhang2 | Jie Shi2 | Ying Yuan1 | Xiaofeng Tao1
1Department of Radiology, Shanghai Ninth People's Hospital, Shanghai JiaoTong University School of Medicine, Shanghai, China | 2MR Research, GE Healthcare, Shanghai, China
Correspondence: Ying Yuan ([yuany83@163.com](yuany83@163.com)) | Xiaofeng Tao ([cjr.taoxiaofeng@vip.163.com](cjr.taoxiaofeng@vip.163.com))
Received: 12 November 2025 | Revised: 6 March 2026 | Accepted: 30 March 2026
Keywords: habitat | lymph node metastasis | radiomics | SyMRI | | | --- | --- | | ABSTRACT
Purpose: To develop an interpretable radiomics model based on habitat for predicting lymph node metastasis in oral cancer using Synthetic MRI (SyMRI) .
Methods: A retrospective study of 101 oral cancer patients with pathologically confirmed lymph node status was performed, dividing them into a training cohort (N = 71) and a test cohort (N = 30) . All patients underwent MRI, including the MAGiC (SyMRI) sequence. Intra-tumoral volumes of interest (VOIs) were clustered into three spatial habitats using the K-means algorithm. Radiomics models were developed for the entire tumor and clustered intra-tumor regions, based on features from synthetic quantitative T1 and T2 maps. The Shapley Additive Explanations (SHAP) method was used to interpret predictions. A radiomics nomogram was constructed by integrating independent variables, including the radiomics score and MRI-reported status. Predictive performance was evaluated using AUC, DeLong's test, calibration curves, and decision curve analysis (DCA) . Results: The habitat with the highest T1 and a moderately low T2 values within the tumor showed the best predictive performance, compared to whole-tumor radiomics (training: 0.87 vs. 0.82; test: 0.74 vs. 0.62) . The radiomics nomogram, combining radiomics features and independent clinical variables, outperformed clinical diagnosis alone in both the training cohort (0.92 vs. 0.80; p = 0.0023) and test cohort (0.84 vs. 0.73; p = 0.0362) .
Conclusion: The habitat-based radiomics signature, integrating SyMRI-derived quantitative T1 and T2 maps, provides enhanced predictive accuracy for preoperative lymph node metastasis in oral cancer compared to clinical diagnosis. | | | 1 | Introduction | diagnoses and 14,1465 deaths in Asia in 2022 [1] . The presence of lymph node metastasis is significantly associated with a poor | | Oral cancers, accounting for approximately 3%–5% of all ma- | prognosis, leading to a 50% reduction in overall survival rates. | | lignancies, are widely acknowledged as one of the most prev- | Performing upfront elective neck dissection (END) can effec- | | alent global malignancies, with an estimated 25,8440 new | tively avoid missed diagnoses, whereas unnecessary END may | | Abbreviations: AUC, area under the curve; DCA, decision curve analysis; DOI, depth of invasion; END, elective neck dissection; ICC, interclass correlation coefficient; LASSO, least absolute shrinkage and selection operator; LightGBM, light gradient boosting machine; LR, logistic regression; MAGiC, magnetic resonance image compilation; NRI, categorical net reclassification improvement; RF, random forest; ROC, receiver operating characteristic; SHAP, Shapley Additive Explanations; SMOTE, synthetic minority over-sampling technique; SVM, support vector machine; SyMRI, synthetic MRI; VOIs, volumes of interest; XGBoost, eXtreme gradient boosting.
Rui Wang and Jiliang Ren have contributed equally to this study and should be considered as co-first authors. | | | This is an open access article under the terms of the Creative Commons Attribution-NonCommercial-NoDerivs License, which permits use and distribution in any medium, provided the original work is properly cited, the use is non-commercia

FAQs

What is the study’s main goal?

To develop an interpretable habitat-based radiomics model using SyMRI-derived quantitative T1 and T2 maps for predicting lymph node metastasis in oral cancer.

How are the “habitats” defined for the radiomics model?

Intra-tumoral VOIs are clustered into three spatial habitats using the K-means algorithm, and radiomics models are built for the whole tumor and for clustered regions.

How is model interpretability achieved?

The Shapley Additive Explanations (SHAP) method is used to interpret the predictions.

Sep 22, 2026

Uploaded by Riley West