Probability-Level Fusion of Meta-Information and Multimodal MRI/PET Radiomics for Predicting 1p/19q Codeletion and IDH Mutation in Gliomas

Noninvasive prediction of 1p/19q codeletion and isocitrate dehydrogenase (IDH) mutation may support preoperative assessment of lower-grade gliomas. We evaluated leakage-aware probability-level fusion of meta-information and multimodal radiomics. This retrospective study included 217 patients with World Health Organization grade II or III gliomas and preoperative magnetic resonance imaging and positron emission tomography. Meta-information and radiomics models were developed separately, and their calibrated probabilities were combined using weighted averaging or logit stacking. Performance was evaluated using repeated stratified five-fold nested cross-validation with 10 seeds, with all data-dependent procedures confined to training data. For 1p/19q codeletion, weighted averaging achieved an outer-fold area under the receiver operating characteristic curve (AUC) of 0.816 ± 0.056 (mean ± standard deviation), and both fusion approaches achieved higher patient-level AUCs than meta-information alone after Holm correction. For IDH mutation, weighted averaging achieved an AUC of 0.799 ± 0.077 and a higher patient-level AUC than radiomics alone, whereas the comparison with meta-information alone was not statistically supported. Modality-group analyses revealed task-dependent performance patterns without identifying a consistently superior modality condition after multiplicity correction. Probability-level fusion showed task-dependent incremental discrimination and warrants further external and prospective evaluation with contemporary molecular reference standards and prespecified imaging harmonization procedures.

Authors

Institutions

Publication Details

Journal
Journal of Imaging
Published
2026-11-13
DOI
https://doi.org/10.3390/jimaging12110572
Primary Topic
Radiomics and Machine Learning in Medical Imaging
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Probability-Level Fusion of Meta-Information and Multimodal MRI/PET Radiomics for Predicting 1p/19q Codeletion and IDH Mutation in Gliomas

神田真理恵, Kaori Kusuda, Yoshihiro Muragaki, Ken Masamune et al.
3 links
Journal of Imaging
Radiomics and Machine Learning in Medical Imaging
article

Probability-Level Fusion of Meta-Information and Multimodal MRI/PET Radiomics for Predicting 1p/19q Codeletion and IDH Mutation in Gliomas

神田真理恵, Kaori Kusuda, Yoshihiro Muragaki, Ken Masamune, Yuji Ohta
article en

Abstract

Noninvasive prediction of 1p/19q codeletion and isocitrate dehydrogenase (IDH) mutation may support preoperative assessment of lower-grade gliomas. We evaluated leakage-aware probability-level fusion of meta-information and multimodal radiomics. This retrospective study included 217 patients with World Health Organization grade II or III gliomas and preoperative magnetic resonance imaging and positron emission tomography. Meta-information and radiomics models were developed separately, and their calibrated probabilities were combined using weighted averaging or logit stacking. Performance was evaluated using repeated stratified five-fold nested cross-validation with 10 seeds, with all data-dependent procedures confined to training data. For 1p/19q codeletion, weighted averaging achieved an outer-fold area under the receiver operating characteristic curve (AUC) of 0.816 ± 0.056 (mean ± standard deviation), and both fusion approaches achieved higher patient-level AUCs than meta-information alone after Holm correction. For IDH mutation, weighted averaging achieved an AUC of 0.799 ± 0.077 and a higher patient-level AUC than radiomics alone, whereas the comparison with meta-information alone was not statistically supported. Modality-group analyses revealed task-dependent performance patterns without identifying a consistently superior modality condition after multiplicity correction. Probability-level fusion showed task-dependent incremental discrimination and warrants further external and prospective evaluation with contemporary molecular reference standards and prespecified imaging harmonization procedures.

Journal of ImagingVol. 12(11)
Kyoto University (JP), Ochanomizu University (JP), Tokyo Women's Medical University (JP), Center for Digital Transformation of Healthcare, Graduate School of Medicine, Kyoto University (JP), Institute of Advanced Biomedical Engineering and Science, Tokyo Women's Medical University (JP), Center for Advanced Medical Engineering Research & Development, Kobe University (JP), Kobe University (JP)
Good health and well-being
Radiomics and Machine Learning in Medical Imaging
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon