MixBranchNet: Glioma Segmentation and Genotype Prediction Data from CEST MRI

MixBranchNet is a dataset supporting a task-adaptive deep learning model for glioma segmentation and genotype prediction using chemical exchange saturation transfer (CEST) MRI. The data includes multi-offset CEST images with 41 frequency offsets from a 3.0 T MRI scanner, with manual annotations from neuroradiologists and genotype labels from histopathology. The dataset, authored by Yanduo Li, was last updated on 2026-04-23 and is shared under a CC-BY-4.0 license.

Use Cases

• Training segmentation models for brain tumors based on multi-offset CEST MRI data.
• Developing genotype prediction models for IDH mutation and MGMT promoter methylation status from imaging features.
• Benchmarking spatial-spectral deep learning architectures against conventional CEST quantification techniques.
• Conducting methodological research on task-adaptive networks for combined segmentation and classification.

Strengths

• Performance metrics are provided, including a Dice coefficient of 0.84 for segmentation and accuracy up to 95.00% for genotype prediction.
• Data is validated against a reference standard from two board-certified neuroradiologists and surgical histopathology.
• Methodology includes five-fold cross-validation with strict patient-level partitioning and a fully isolated hold-out test set.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• The description indicates a single-center cohort, which may limit generalizability.

Provenance

Source

figshare, authored by Yanduo Li.

Collection Method

CEST MRI images from a 3.0 T scanner, with manual annotations from neuroradiologists and genotype labels from histopathology.

Time Range

null

Freshness

Last updated 2026-04-23 04:20:52; freshness should be verified.

Geography

null