Related Experiment Video
Updated: Oct 2, 2026
A Cognitive Fusion-guided Prostate Biopsy Using Multiparametric Magnetic Resonance Imaging and Transrectal Ultrasound
Published on: March 21, 2025
Technical and biological validation of prostate apparent diffusion coefficient measurements on an MRI-linac
Damien J McHugh1, Martin Swinton2, Claire A Hart3
1Christie Medical Physics and Engineering, The Christie NHS Foundation Trust, Manchester, United Kingdom; Division of Cancer Sciences, The University of Manchester, Manchester, United Kingdom.
Purpose:
Quantitative MRI biomarkers require validation before being used to assess treatment response or inform biological image-guided adaptive radiotherapy on MRI-linacs. This work contributes to the technical validation of prostate apparent diffusion coefficient (ADC) measurements by evaluating phantom accuracy, intra-scanner repeatability, and inter-scanner reproducibility, and contributes to biological validation by evaluating correlations with histology.
Methods:
Phantom ADC was measured on a 1.5 T MRI-linac (MRL) and 1.5 T diagnostic MR (dMR). ADC in non-lesion tissue (ADCNL) and in the dominant intraprostatic lesion (ADCDIL) was measured twice in 16 patients on MRL and/or dMR before prostatectomy. Repeatability was quantified through Bland-Altman analysis and within-subject coefficients of variation (wCV). Reproducibility was quantified through Bland-Altman analysis and comparison with diffusion microstructural model simulations. Tissue samples were analysed using haematoxylin and eosin (H&E) and immunohistochemistry (IHC). Spearman's ρ quantified relationships between ADC and histology-derived epithelial (EF) and stroma (SF) fraction.
Results:
Phantom mean bias was 0.024 (MRL) and -0.019 (dMR) µm2/ms. Patient ADCs pooled across MRL and dMR exhibited excellent repeatability: wCVNL = 1.4%, wCVDIL = 4.6%. ADCNL was higher on MRL than dMR: mean bias [95% confidence interval] = 0.20 [0.14, 0.26] µm2/ms, consistent with simulations. Correlations between ADCDIL and histology-derived metrics were consistent in magnitude and direction for both scanners. Pooled data showed ADCDIL and EFH&E were negatively correlated (ρ = -0.69, p = 0.003); ADCDIL and SFH&E were positively correlated (ρ = 0.70, p = 0.003).
Conclusion:
Systematic differences exist between MRI-linac and diagnostic MR measurements. ADCDIL from both systems showed similar correlations with tissue microstructure. This study contributes to the technical and biological validation of diagnostic MR and MRI-linac prostate ADC.
