Nuclear Medicine

Most commercial radiology AI validations omit subgroup performance reporting.

This scoping review evaluated 545 peer-reviewed validation studies of commercial radiology AI products published after 2010. Only 392 studies reported any demographic subgroup data and just 77 provided subgroup performance results, impeding bias assessment. Reporting varied by modality and image type, with skeletal and lung studies relatively more likely to present subgroup data. The authors found no temporal or sponsorship-related improvement in subgroup reporting and noted many tuberculosis datasets were underpowered. 1

Nomogram using serum thyroglobulin and IFN-α predicts response to initial RAI in intermediate/high-risk DTC.

This retrospective study of 429 patients developed a nomogram predicting 6–12 month response to initial radioactive iodine (RAI). In multivariable analysis, lower serum thyroglobulin (sTg) and higher IFN-α independently predicted an excellent response (ORs reported in the training cohort). The model showed acceptable discrimination and calibration in internal and external validation cohorts. Decision curve analysis suggested potential clinical utility for individualized risk stratification and follow-up planning. 2

Two-fraction, dosimetry-guided 177Lu-DOTATATE increases tumour dose while maintaining organ safety in neuroblastoma.

In 14 high-risk recurrent/refractory neuroblastoma patients, two fractions of 177Lu-DOTATATE were given 2–3 weeks apart with dosimetry guidance. Tumour-to-kidney and tumour-to-whole-body dose ratios fell by 44% after the second fraction, and tumour dose per activity decreased by median 53%. Kidney and whole-body doses per activity decreased by 24% and 18%, respectively, and no sustained renal toxicity was reported. Four of 14 patients required peripheral blood stem cell reinfusion, prompting protocol amendment to front-load activity to 400 MBq/kg for subsequent patients. 3