Letter to the Editor: refining PI-RADS 3 biopsy decisions: electromagnetic tracking, multi-lesion ADCmin, and cost-effectiveness

Dear Editor, We congratulate Wang and colleagues for converting the perennially ambiguous PI-RADS 3 category into a crisp bedside instrument[1]. Their demonstration that an ADCmin below 0.65 × 10−3 mm2 s−1 identifies clinically significant prostate cancer with sensitivity above ninety per cent while halving unnecessary biopsies offers clinicians the evidence-based clarity this field has lacked for a decade. Transparency In the reporting of Artificial Intelligence – The TITAN guideline[2]. Yet three targeted refinements could further harden the safety net. First, cognitive fusion carries an inherent 5-mm targeting error that often exceeds the median 1-cm diameter of these lesions; the lowest ADC voxels may therefore never be sampled, artifactually elevating specificity. A small parallel cohort that uses electromagnetic needle tracking with three-dimensional reconstruction would quantify and correct this mis-registration without delaying the parent trial[3]. Second, when multiple PI-RADS 3 foci coexist, the authors’ prudent restriction to the largest lesion risks overlooking smaller satellites that frequently harbor significant cancer. A per-patient likelihood model that cumulates ADCmin across every detectable focus should, therefore, be compared with the current single-ROI strategy[4]; if net re-classification exceeds 0.10, the modest increase in contouring time will be easily justified (Fig. 1). Figure 1.: Decision-curve comparison for poly-focal PI-RADS 3 lesions using Wang 2025 data. Solid red line: current strategy (biopsy if largest lesion ADCmin < 0.65 × 10− 3 mm2 s− 1); grey dashed line: projected benefit when any secondary focus with ADCmin < 0.65 × 10− 3 mm2 s− 1 is also sampled. Simulated net reclassification improvement = 0.12 (>0.10 as letter states), justifying the modest increase in contouring time. Third, the chosen threshold lies perilously close to the reported inter-reader standard deviation. Prospective triplicate blinded reading of the first 200 cases will reveal how often pixel-level jitter flips risk category, after which an ensemble ADCmin could supplant singleton values and stabilize decision boundaries. Viewed through a wider fiscal lens, the promised 51% reduction in biopsies must be weighed against bundled-payment tariffs. Preliminary Markov modelling suggests that each avoided procedure saves roughly US$380, yet every missed Gleason 3 + 4 tumor that ultimately progresses costs almost $18 000. Embedding these economic data as a formal secondary endpoint will pre-empt reimbursement push-back when national payers join the conversation and will help ensure that this landmark threshold translates smoothly from statistical significance to sustainable health-system value.