Whole-body diffusion-weighted imaging (WB-DWI) is increasingly used in staging and therapy assessment for metastatic prostate cancer. The authors evaluated whether an accelerated whole-body diffusion-weighted imaging acquisition reconstructed with a research deep learning (DL) variational-network package could reduce scan time while maintaining perceived image quality and quantitative ADC estimates of lesions.
The primary objective was to compare image quality between the standard WB-DWI protocol and an accelerated DL-reconstructed WB-DWI protocol. A secondary objective was to assess agreement in lesion ADC values between the two protocols.
This was a single-centre prospective study performed at 1.5T. Two patient cohorts were recruited under a registered clinical trial (NCT05118555) with ethical approval from a national research ethics committee (London - Central Research Ethics Committee, REC reference 21/LO/0079).
Cohort 1 (n = 10) underwent both standard and accelerated DL WB-DWI for qualitative image-quality assessment. Cohort 2 (n = 20) underwent both protocols for quantitative ADC comparison in lesions identified as hypercellular focal bone metastases by a radiologist. The report does not provide additional demographic or clinical details beyond cohort sizes and setting.
The accelerated WB-DWI acquisition used fewer signal averages and higher parallel imaging acceleration relative to the centre’s standard-of-care WB-DWI. Images from the accelerated acquisition were processed with a research DL reconstruction package based on a variational network. The DL approach was applied to the accelerated acquisitions to produce images for reader assessment and ADC measurement.
The authors report a 37% reduction in whole-body DWI acquisition time with the accelerated DL WB-DWI protocol compared with the standard protocol.
In Cohort 1, two radiologists performed a blinded qualitative assessment of image quality across four anatomical stations. Image quality was scored using a 4-point Likert scale. Readers were blinded to which DWI protocol produced each dataset.
Radiologists consistently rated both the standard and the accelerated DL WB-DWI protocols as good-to-excellent across all image-quality metrics. The accelerated protocol did not show perceived degradation in image quality in these blinded assessments. Image-quality scores were analysed using Wilcoxon signed-rank tests with Bonferroni correction, as reported by the authors.
In Cohort 2, the investigators evaluated apparent diffusion coefficient (ADC) values measured in hypercellular focal bone metastases identified by a radiologist. ADC values from the standard and accelerated DL WB-DWI protocols were compared.
Agreement between ADC measurements from the two protocols was assessed using Bland–Altman analysis. The authors report that median ADC estimates in lesions showed no significant difference between standard and accelerated DL WB-DWI. Similarly, the interquartile range of ADC estimates did not differ significantly between protocols according to the presented analysis.
The authors conclude that DL-based reconstruction enables a 37% reduction in WB-DWI acquisition time without degrading image quality or affecting lesion ADC quantification in patients with metastatic prostate cancer. They suggest these findings support clinical implementation of accelerated DL WB-DWI for staging and therapy assessment in this population.
As this work is reported as a preprint, the authors and medRxiv note the study has not been peer reviewed and should not yet be used to change clinical practice until further validation and peer review are completed.
Ethical approval was provided by a national research ethics committee (London - Central Research Ethics Committee; REC reference 21/LO/0079). The study is registered with ClinicalTrials.gov (NCT05118555). The authors declare no competing interests.
Funder information reported includes independent research funding from the NIHR Biomedical Research Centre at The Royal Marsden NHS Foundation Trust and The Institute of Cancer Research, Royal Marsden Cancer Charity, Prostate Cancer UK and Cancer Research UK National Cancer Imaging Trials Accelerator (NCITA).
All data produced in the study are available upon reasonable request to the authors. The manuscript is presented as a preprint on medRxiv and has not been certified by peer review.
Note: The abstract and article metadata provide the results and methods summarized above. Specific numeric image-quality scores, detailed ADC values, full reader agreement statistics, and protocol parameter tables are not reported in the abstract and should be consulted in the full text or supplementary material for comprehensive appraisal.