The global MRI community this week traveled to paradise to convene its annual meeting of the International Society for Magnetic Resonance in Medicine. If you were one of the lucky ones to be in attendance in Honolulu, Hawaii for ISMRM 2025, you were treated to some of the latest news in radiology’s most powerful modality.
As has been the case at other radiology meetings, AI took center stage in Honolulu.
- AI has multiple use cases in MRI, from helping radiologists interpret images more efficiently to accelerating scans and upscaling lower-field images to resemble high-field exams.
Just a few of the news highlights from ISMRM 2025 are below …
- Using AI to interpret prostate MRI reduced reading times by 48% (250 to 120 seconds) while improving the diagnostic performance of both experienced and less experienced radiologists.
- AI of thyroid T2-weighted neck MRI scans demonstrated good accuracy (87%) for nodules larger than 1 cm, indicating a possible role for screening and monitoring.
- Researchers presented progress in creating brain charts of white matter based on MRI scans of 24k cognitively healthy people that can be used to track normal and abnormal brain development.
- Brain MRI showed that lower brain volumes in people with coronary artery disease were associated with worse aerobic fitness and higher BMI, revealing a link between cardiovascular and brain health.
- Chinese researchers showed their work on PMEEN, a multimodality brain scanner that combines PET, MRI, EEG, eye-tracking, and functional near-infrared spectroscopy.
- A Spanish team demonstrated research on a low-field PET/MRI scanner with focused ultrasound capability for therapeutic applications.
- AI could be used during abbreviated breast MRI screening scans to convert women mid-exam to a full MRI protocol if abnormalities are detected.
- 7T MRI was used to detect iron deposits in the brain, which could be a marker for Alzheimer’s disease.
- MRI with an ultrashort echo time protocol could be an alternative to CT for following up lung nodules.
- Researchers presented a deep learning-based approach to generating synthetic contrast-like MR images without gadolinium.
- MGH researchers showed progress in developing a 136mT portable MRI scanner for bedside brain scanning of preterm neonates.
The Takeaway
The rapid proliferation of news about AI-based MRI at ISMRM 2025 suggests its own vision of paradise – a world in which MRI can be deployed more widely than ever before, where radiologists with AI assistance detect disease in many cases before symptoms even occur. We can only dream.
