Noninvasive Risk Stratification of Prostate Cancer Patients Using Radiomic Features Derived from Magnetic Resonance Fingerprinting (MRF) and MRI
Abstract
Objective and Rationale: Prostate cancer is the sole malignancy that uses blind biopsy procedures to confirm disease presence and assign appropriate treatment options based on the risk of disease progression (low, intermediate, and high). A prostate biopsy involves drawing a number of random samples of prostate gland using needles, a procedure causing immense pain, discomfort, rectal bleeding, and risk of infections in prostate cancer patients. Prostate biopsies are found to miss cancer in ~50% of patients, and ~75% of men who undergo prostate biopsy are found to have no cancer. A recent study also revealed an increase in hospitalization, as well as death rate, in older patients and patients with other diseases following prostate biopsy. There is therefore a clear unmet need to (1) develop a non-invasive method of identifying patients with no prostate cancer who in turn do not need to obtain a biopsy; (2) identify low-risk prostate cancer patients who are potential candidates for active surveillance (watchful waiting) who would otherwise undergo radical therapy and suffer from related complications (impotence, incontinence, rectal bleeding); and (3) develop technology that could allow for non-invasive monitoring of these patients on active surveillance. Multi-parametric magnetic resonance imaging (mpMRI), which is a non-invasive modality, is now widely used in prostate cancer diagnosis, but is limited by inter-reader differences in interpreting MRI (missing ~10% of aggressive prostate cancers) and similar appearance of cancerous and benign or non-cancerous (~60% of suspicious low-risk cancers are benign) lesions. To minimize differences in interpreting MRI, magnetic resonance fingerprinting (MRF) technology has been recently developed to obtain reliable and accurate quantitative maps. Preliminary results using MRF has shown that nearly accurate separation between cancer and normal prostate is possible; however, benign tumors appearing as cancerous could not be well separated. Radiomics refers to computational methods that capture subtle and sub-visual patterns in images that are not visible on visual inspection. The Principal Investigator’s (PI’s) previous work has shown that radiomic features, such as Haralick, Gabor, and co-occurrence of local anisotropic gradient orientations (CoLlAGe) that capture subtle microarchitectures and image gradients, can help distinguish low- and high-risk prostate cancers and also predict recurrence in patients undergoing surgery. In the context of MRF, recent preliminary results generated by the PI show improved separation between cancer and prostatitis on N=39 patients, using radiomics on MRF. Our hypothesis is that employing radiomic features derived from MRF, along with mpMRI, which has a better anatomic detail, could potentially result in discovering useful imaging signatures that distinguish different grades of cancer lesions while also distinguishing them from benign lesions, leading to developing a non-invasive and accurate method for prostate cancer risk assessment. Contributions and Applicability: The major contributions in the proposed study are as follows: 1. Developing and implementing novel radiomic approaches in the context of MRF to achieve non-invasive risk stratification of prostate cancer patients. Radiomic features on MRF that characterize and distinguish aggressive and indolent cancer lesions can help specifically identify patients with aggressive disease and allow for early and aggressive treatment. This will ultimately lead to minimizing mortality due to disease progression. 2. Obviating the need for biopsies in men with elevated levels of prostate specific antigen (PSA) and therefore reducing the side effects associated with biopsies. This will further allow for identifying and retaining men who are candidates for active surveillance and allow for non-invasive monitoring of these patients. 3. Radiomic features identified in this project w
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Oct 29, 2018
- Source ID
- W81XWH1810524
Entities
People
- Rakesh Shiradkar
Organizations
- Case Western Reserve University
- United States Army