ASLNet: ASL MRI Scans for Glioma Grading and Survival Prediction

471 patients with histologically confirmed diffuse glioma underwent ASL MRI between 2015 and 2021. The dataset was used to train the ASLNet framework, an interpretable three-dimensional residual network, for predicting WHO tumor grade and overall survival. The study was authored by Rafail C. Christodoulou and shared under a CC-BY-4.0 license.

Use Cases

• Training glioma grading models based on noninvasive ASL MRI perfusion information.
• Developing survival prediction models based on ASL imaging fused with demographic and clinical variables like age and extent of resection.
• Generating saliency maps to identify influential perfusion regions in tumors for model interpretability.

Strengths

• Includes data from 471 patients, providing a substantive sample size for model development.
• Model performance metrics are reported, including an AUC of 0.79 for grading and 0.70 for survival prediction.
• Data is shared under a permissive CC-BY-4.0 license, facilitating reuse.

Limitations

• The dataset is very small at 46.1 KB, suggesting it contains documentation or a model description rather than the raw image data.
• Column-level documentation is absent; field semantics must be inferred after download.
• Data may reflect temporal bias inherent to the single-institution, retrospective collection from 2015-2021.

Provenance

Source

figshare

Collection Method

Retrospective study using original three-dimensional ASL MRI volumes from patients with diffuse glioma.

Time Range

2015 to 2021

Freshness

Last updated 2026-05-18 05:43:08; freshness should be verified.