Back

Post-hoc Estimation of a Quantitative Restriction Spectrum Imaging Biomarker for Prostate Cancer Detection Using Conventional MRI

Do, D.; Conlin, C. C.; Bagrodia, A.; Cooperberg, M.; Hahn, M. E.; Harisinghani, M.; Hollenberg, G.; Javier-Desloges, J.; Kamran, S.; Kane, C. J.; Lee, K.-L.; Liss, M. A.; Margolis, D. J.; Murphy, P.; Nakrour, N.; Ohliger, M.; Osinski, T.; Rakow-Penner, R.; Rojo Domingo, M.; Salmasi, A.; Shabaik, A.; Song, Y.; Trecarten, S.; Wehrli, N.; Weinberg, E.; Woolen, S.; Dale, A. M.; Seibert, T. M.

2025-10-15 oncology
10.1101/2025.10.10.25337687 medRxiv
Show abstract

BackgroundMultiparametric MRI is useful for early detection of clinically significant prostate cancer (csPCa), but its standard Apparent Diffusion Coefficient (ADC) has limited utility as a quantitative metric for automated, patient-level detection of csPCa. Restriction Spectrum Imaging (RSI), an advanced diffusion technique, yields a quantitative biomarker (RSIrs) that improves csPCa detection. RSIrs is typically calculated from a dedicated multi-b-value acquisition. RSIrs estimated from conventional MRI has not been studied. PurposeTo evaluate the accuracy and validity of RSI metrics estimated post-hoc from conventional diffusion-weighted imaging (DWI) to serve as a viable surrogate for a dedicated RSI acquisition. Materials and MethodsWe conducted a retrospective, multi-center study of patients with both a dedicated RSI acquisition and conventional DWI. We compared three different RSI restriction score (RSIrs) calculation methods: from the dedicated acquisition (RSIrsdedicated), from conventional DWI alone (RSIrspost-hoc), and from a combination of conventional DWI with only the high b-values from the RSI acquisition (RSIrscombo). We compared these methods for quantitative agreement and csPCa detection performance (Area under the Receiver Operating Characteristic [AUC, 95% Confidence Interval]) of maximum RSIrs (RSIrsmax) in the prostate compared to that of minimum ADC (ADC). ResultsData from n=1095 patients (16 centers) were analyzed. Post-hoc RSIrsmax differed systematically from RSIrsdedicated by a median of +156 (RSIrspost-hoc) and -59 (RSIrscombo), respectively. AUCs for csPCa detection were 0.51 [0.47,0.54], 0.60 [0.57,0.64], 0.70 [0.67,0.74], and 0.77 [0.74,0.80] for ADC, RSIrspost-hoc, RSIrscombo, and RSIrsdedicated, respectively. ConclusionEven when estimated using conventional DWI, RSIrs is a superior quantitative biomarker to ADC for automated, patient-level detection of csPCa. A dedicated RSI acquisition gives the best performance. A compromise would be to acquire high b-values (1500 s/mm2, 2500 s/mm2) to complement low b-values (<1000 s/mm2) from conventional DWI.

Matching journals

The top 8 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.