30-Second Takeaway
- Photon-counting CT enables substantial radiation and iodine reductions while improving lung cancer lesion conspicuity and diagnostic confidence.
- PCD-CT supports systematic iodinated contrast dose reduction using a simple 10%-per-5-keV virtual monoenergetic rule.
- Interpretable AI tools across breast, pancreas, head CT, and liver offer targeted sensitivity gains but require workflow and triage planning.
- AI-augmented structured reporting improves chest radiograph accuracy and efficiency compared with free-text reporting.
- Post-ablation cT1 RCC surveillance CT is likely overused, with no evident survival benefit from more intensive schedules.
Week ending February 7, 2026
PCCT, AI decision support, and imaging stewardship: tightening signal, cutting noise
Photon-counting CT lowers dose and improves lung cancer enhancement assessment
In this prospective study of 200 matched lung cancer patients, low-dose PCCT reduced effective dose by about two-thirds versus standard EID CT (1.36 vs 4.04 mSv). PCCT also cut iodine load by roughly one-quarter and was associated with fewer adverse reactions and contrast-induced acute kidney injury events. Ultrahigh-resolution 0.4-mm PCCT sections improved detection of enhancement-related malignant features and raised diagnostic confidence compared with EID CT. Image quality gains were most pronounced for normal BMI patients and lesions ≤3 cm, supporting PCCT use in early-stage (T1) lung cancer workup.
The PCCT "10-to-5 rule" for iodinated contrast reduction
Phantom and retrospective patient data on dual-source PCD-CT were used to link virtual monoenergetic image energy with iodine contrast-to-noise ratio (CNR). Lowering VMI energy from 60 to 40 keV increased CNR at a given iodine concentration, largely independent of phantom size, dose, and kV. Across abdominal portal-venous and CTA protocols, each 5-keV VMI reduction supported about 10% iodinated contrast dose reduction while maintaining CNR. Patient data confirmed similar CNR gains per 5-keV step, yielding a practical "10%-per-5 keV" rule to guide PCCT contrast protocol optimization.
Interpretable MRI AI cuts false positives in BI-RADS 4 breast lesions
BL4AS, an interpretable AI system for BI-RADS 4 breast MRI lesions, was trained on 2,803 lesions from 2,686 women. The model achieved AUCs of 0.892–0.930 and markedly higher specificity than radiologists (0.889 vs 0.491). AI-assisted reading improved diagnostic accuracy for both junior and senior radiologists and reduced inter-reader variability by nearly one quarter. False-positive rates fell by 27.3%, and the system provided BI-RADS 4 sub-stratification, supporting more individualized biopsy and follow-up decisions.
CAD detects pancreatic cancer on diagnostic and prediagnostic CT with high AUC
This multicenter CAD tool for pancreatic cancer was trained on 2,496 contrast-enhanced CTs and externally validated at two institutions. On diagnostic test sets including 200 cancers and nearly 5,000 controls, the tool achieved AUC 0.95 with 90% sensitivity and 87.8% specificity. For prediagnostic CTs obtained 1–12 months before clinical diagnosis, sensitivity was 66.7%, reflecting detection of many clinically missed tumors. Sensitivity for tumors ≤2 cm remained substantial in both diagnostic and prediagnostic sets, suggesting potential for earlier detection of small cancers.
References
Numbered in order of appearance. Click any reference to view details.
Additional Reads
Optional additional studies from this edition.