Deep learning with quantitative MRI enables noninvasive treatment planning and molecular profiling of adult diffuse gliomas.
Key Details
- 1Study involved 1,008 adults with confirmed diffuse gliomas.
- 2Researchers used six co-registered quantitative MRI maps including ADC, CBF, T1, T2, PD, and contrast-enhanced T1.
- 3Participants were divided into a training group, an internal validation group, and three external test cohorts.
- 4The team developed five model families: VOI-mean, radiomics, and DenseNet-201 deep learning models (2D, 2.5D, and 3D).
- 5One single-map model achieved the best external test AUC for presurgical prediction of isocitrate dehydrogenase mutation, outperforming others.
Why It Matters
This research highlights the potential of combining advanced MRI techniques with deep learning for accurate and individualized management of diffuse gliomas. It underscores growing opportunities for radiologists to employ AI-driven tools in noninvasive diagnostic and treatment decisions.
Source
AuntMinnie
Related News
•AuntMinnie
DeepHealth AI Suite Expands Radiology Insights Across Modalities
DeepHealth's Radiology AI Suite enables radiologists to detect more findings and extract additional risk and disease progression information from a wide range of imaging exams.
•AuntMinnie
Majority of Radiologists Use AI Daily but Have Mixed Views on Benefits
Survey shows 64% of radiologists use AI daily, but only a minority feel it significantly enhances skills.
•AuntMinnie
AI Boosts Sensitivity but Raises False Alarms in Microbleed Detection
AI significantly increases sensitivity for detecting cerebral microbleeds on brain MRI but at the cost of more false positives, highlighting the complementary strengths of AI and radiologists.