30-Second Takeaway
- PSA density thresholds can safely streamline MRI use and biopsies in MRI-directed prostate cancer pathways.
- CT-based deep learning opportunistically predicts vertebral fractures better than FRAX using existing abdominal scans.
- MRI ENZIAN/#ENZIAN classification is reliable for key deep endometriosis compartments and supports standardized reporting.
- Photon-counting CT and DL-based contrast reduction define new pediatric and neuro MRI dose benchmarks.
- Emerging PET and lung CT AI tools refine lesion detection, localization, and workflow without additional patient burden.
Week ending December 6, 2025
Radiology AI, protocol optimization, and risk-stratified imaging: updates with near-term impact on daily practice
PSA density thresholds refine MRI indication and biopsy decisions in MRI-directed prostate cancer workups
In this post hoc analysis of 996 biopsy-naïve, MRI-pathway men from MR-PROPER, GG ≥2 cancer prevalence was 24%. Using a post-MRI PSA density ≥0.20 ng/ml² in PI-RADS 3 lesions missed only 1.3% of GG ≥2 cancers while avoiding 29% of negative biopsies and some GG1 detections. Raising the post-MRI PSA density threshold from 0.10 to 0.20 therefore substantially improves the biopsy benefit–harm ratio in equivocal MRI findings. A pre-MRI PSA density ≥0.10 ng/ml² could avoid about 30% of MRIs and many unproductive biopsies, but at the cost of nearly 10% missed GG ≥2 cancers, indicating narrow safety margins.
Multitask CT deep learning using bone and muscle improves vertebral fracture risk prediction over FRAX
This study developed a multitask deep learning model on abdominal CT scans from 2553 patients aged 50–80 years, with external validation in 1506 patients. For prevalent vertebral fracture detection, the bone+muscle model achieved AUROC 0.82 in development and 0.80 externally. For 2-, 3-, and 5-year vertebral fracture prediction, multitask bone+muscle models outperformed bone-only models and achieved AUROCs up to 0.79. The multitask image model also modestly outperformed FRAX (c-index 0.68 vs 0.66), suggesting value for opportunistic fracture risk stratification from routine CT.
MRI ENZIAN/#ENZIAN classification shows high accuracy for key deep endometriosis compartments
This meta-analysis pooled 12 studies with 1024 patients to evaluate MRI-based ENZIAN/#ENZIAN versus surgery for deep endometriosis. MRI ENZIAN/#ENZIAN showed high reliability and diagnostic performance for ovaries (O), tubo-ovarian condition (T), vagina/rectovaginal space (A), uterosacral ligaments (B), and rectosigmoid (C). Sensitivity/specificity were about 98%/96% for O and 94%/89% for T, with slightly lower but still high values for A, B, and C. Diagnostic performance for adenomyosis, bladder, and intestinal compartments was lower, and data were insufficient for some pelvic sites, limiting confident staging there.
References
Numbered in order of appearance. Click any reference to view details.
Additional Reads
Optional additional studies from this edition.