Preoperative prediction of Ki-67 proliferation index in gliomas using multisequence and multi-regional radiomics with machine learning and deep learning integration

The Ki-67 proliferation index is an important biomarker for evaluating glioma aggressiveness but is conventionally determined by postoperative histopathological examination. We developed and internally validated a non-invasive framework integrating multisequence magnetic resonance imaging (MRI) radiomics, deep learning (DL) features, and clinical variables for preoperative prediction of Ki-67 expression in gliomas. A total of 179 patients with pathologically confirmed primary glioma were retrospectively enrolled and randomly assigned to training ( n = 125) and internal validation ( n = 54) cohorts. Handcrafted radiomics features were extracted from intratumoral and multi-scale peritumoral regions on T1-weighted contrast-enhanced, T1FLAIR, and T2-weighted MRI, while high-level imaging features were generated using pre-trained convolutional neural networks. Selected imaging components were integrated with clinical variables to construct multimodal predictive models. The inclusion of peritumoral features showed potential value for improving radiomics-based prediction compared with intratumoral features alone. The multimodal integrated model achieved AUCs of 0.869 (95% CI, 0.803–0.934) and 0.849 (95% CI, 0.744–0.953) in the training and internal validation cohorts, respectively, showing a higher observed AUC than the conventional machine learning, DNN-based, and transformer-based models. The clinical model achieved an internal validation AUC of 0.853. These findings highlight the potential value of integrating multi-regional MRI features for preoperative Ki-67 prediction and provide a basis for further investigating the contribution of peritumoral imaging information to glioma characterization.

Authors

Institutions

Publication Details

Journal
Frontiers in Human Neuroscience
Published
2026-11-19
DOI
https://doi.org/10.3389/fnhum.2026.1984518
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

Preoperative prediction of Ki-67 proliferation index in gliomas using multisequence and multi-regional radiomics with machine learning and deep learning integration

Anke Zhang, Kankai Wang, Xin Zhang, Haoliang Zhu et al.
Frontiers in Human Neuroscience
Radiomics and Machine Learning in Medical Imaging
article

Preoperative prediction of Ki-67 proliferation index in gliomas using multisequence and multi-regional radiomics with machine learning and deep learning integration

Anke Zhang, Kankai Wang, Xin Zhang, Haoliang Zhu, Hangzhe Sun, Bowen Wang, Tianyun Zhang, Mingjie Wang, Yichang Hao, Bowen Shang, Shuyao Chen, Yuanbo Pan, Zizhou Huang, Yi Wang, Xiaoying Du, Yuqi Gu, Lei Li
article en

Abstract

The Ki-67 proliferation index is an important biomarker for evaluating glioma aggressiveness but is conventionally determined by postoperative histopathological examination. We developed and internally validated a non-invasive framework integrating multisequence magnetic resonance imaging (MRI) radiomics, deep learning (DL) features, and clinical variables for preoperative prediction of Ki-67 expression in gliomas. A total of 179 patients with pathologically confirmed primary glioma were retrospectively enrolled and randomly assigned to training ( n = 125) and internal validation ( n = 54) cohorts. Handcrafted radiomics features were extracted from intratumoral and multi-scale peritumoral regions on T1-weighted contrast-enhanced, T1FLAIR, and T2-weighted MRI, while high-level imaging features were generated using pre-trained convolutional neural networks. Selected imaging components were integrated with clinical variables to construct multimodal predictive models. The inclusion of peritumoral features showed potential value for improving radiomics-based prediction compared with intratumoral features alone. The multimodal integrated model achieved AUCs of 0.869 (95% CI, 0.803–0.934) and 0.849 (95% CI, 0.744–0.953) in the training and internal validation cohorts, respectively, showing a higher observed AUC than the conventional machine learning, DNN-based, and transformer-based models. The clinical model achieved an internal validation AUC of 0.853. These findings highlight the potential value of integrating multi-regional MRI features for preoperative Ki-67 prediction and provide a basis for further investigating the contribution of peritumoral imaging information to glioma characterization.

Frontiers in Human NeuroscienceVol. 20
Wenzhou Medical University (CN), First Affiliated Hospital of Bengbu Medical College (CN), Second Affiliated Hospital of Zhejiang University (CN), Zhejiang University (CN)
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