Hybrid Multidimensional Prostate MRI: A Quantitative Step in the Right Direction

See also the article by Lee and Chatterjee et al in this issue.

Dr Baris Turkbey is a senior clinician at the Molecular Imaging Branch (MIB), National Cancer Institute, National Institutes of Health. He currently serves as the head of MRI and Artificial Intelligence Resource at MIB. He is a fellow of Society of Abdominal Radiology and serves as a member of the PI-RADS Steering Committee. His main research areas include cancer imaging (MRI, PET/CT), quantitative imaging biomarkers, and artificial intelligence.

Prostate MRI has been in use for almost 2 decades for guiding prostate biopsies, and its benefit in diagnosing clinically significant prostate cancers with reduced diagnosis of clinically indolent prostate cancers has been reported in large clinical studies (1,2). Along with the growing interest to implement prostate MRI into localized prostate cancer diagnosis, a need for standardized acquisition and interpretation for prostate MRI scans emerged and Prostate Imaging Reporting and Data System (PI-RADS) guidelines were released in late 2014 (3). Currently, its updated version (PI-RADS version 2.1) is in use to further improve the standardization process for prostate MRI scans (4). PI-RADS uses a five-category system for T2-weighted and diffusion-weighted MRI pulse sequences and has a dominant-sequence approach for each zone of the prostate. Imaging features of each category are detailed for each pulse sequence in the hopes that radiologists will have improved intra- and interreader consistency when they read prostate MRI scans.

Unfortunately, research results on the reproducibility of PI-RADS indicate that there is a substantial amount of reader variation when radiologists use the system. This is likely due to subjective and nonquantitative nature of PI-RADS category definitions (5). Ultimately, the low reproducibility of PI-RADS also impacts prostate cancer detection rates among different centers (6).

Quantification of imaging studies certainly enables radiologists to have a more consistent performance for tasks such as lesion detection and classification. CT and PET are among the very successful examples of robustly quantifiable imaging studies, which are commonly used in oncologic radiology. For prostate MRI, there has been substantial research effort to perform and optimize quantifiable pulse sequences such as MR spectroscopy, apparent diffusion coefficient (ADC) mapping of diffusion-weighted MRI, dynamic contrast-enhanced MRI, and T2 mapping for intraprostatic lesion detection and classification tasks (7). While promising results have been reported for some of these pulse sequences (eg, ADC maps) (8), currently none of these approaches is recommended to be used quantitatively in major prostate MRI guidelines (4).

In this issue of Radiology, Lee and Chatterjee et al (9) compare the performance, interobserver agreement, and interpretation time of prostate multiparametric MRI (mpMRI) and hybrid multidimensional MRI (HM-MRI) in diagnosing clinically significant prostate cancer. HM-MRI is a quantitative MRI method that provides tissue percentage estimates based on MRI data by exploiting the interdependence of measured ADC and T2 values (10). HM-MRI data evaluation uses a compartmental prostate tissue model to measure volume fractions of the lumen, epithelium, and stroma of the prostate tissue, as well as the ADC. T2 values are associated with each volume fraction. In this retrospective study with 61 patients who underwent mpMRI and HM-MRI with subsequent prostatectomy or targeted biopsy, four readers with varying experience (1–20 years) evaluated the two MRI techniques to detect prostate cancer lesions, with a 4-week washout period. Patient-level analysis revealed higher or similar area under the receiver operating characteristic curve for HM-MRI compared with mpMRI in three of four readers, and improvement was reported for the least experienced reader (0.64 vs 0.46). While HM-MRI had lower or similar sensitivity for all readers, the specificity of HM-MRI was higher in three of the readers when compared with mpMRI. Additionally, HM-MRI had higher interobserver agreement than mpMRI (Cronbach alpha, 0.88 vs 0.26) and evaluation time for HM-MRI was lower compared with that for mpMRI (73 vs 254 seconds).

Information about tissue composition obtained from prostate MRI not only enables detection of cancer-suspicious foci, but it can also assist prediction of prostate cancer aggressiveness. Traditionally, ADC values have been reported to negatively correlate with tumor Gleason scores; however, their variation and error rate limit their routine use (7). HM-MRI uses quantitative signal features from two pulse sequences and aims to predict lumen, epithelium, and stroma compositions of the prostate tissue. These tissue features are also used to detect and categorize prostate cancers during histopathologic evaluations. Therefore, HM-MRI has a good potential to predict prostate tissue composition in a quantitative fashion. Cancer detection and interobserver agreement results of their retrospective study are quite promising in using a quantitative MRI method for prostate cancer diagnosis.

The study by Lee and Chatterjee et al (9) has some limitations. It is a retrospective study with a relatively small sample size, which does not allow detailed analyses such as zonal lesion detection performance. The study sample does not include a true-negative control group, and the readers are from the same institution where the imaging research was conducted, which might result in an inevitable recall bias.

In summary, quantitative prostate MRI methods have an important potential to enable objective, consistent performance for lesion detection and characterization tasks during radiology readouts. Lee and Chatterjee et al report promising diagnostic performance metrics and interreader agreement for HM-MRI in comparison with mpMRI. The HM-MRI pulse sequence can further assist the current efforts for optimizing prostate MRI acquisition and interpretation. Like every other new imaging technique, HM-MRI needs further evaluation in prospective multicenter studies that include scanners from multiple vendors.