TARGET AUDIENCE

Physicians, imaging scientists/engineers, technologists, and other health professionals who are interested in performing MR molecular imaging of the prostate

OBJECTIVES

This talk will give a brief description of recent advances in molecular imaging techniques such as MRSI and PET/MRI, and how these techniques in combination with Big Data strategies and machine learning may contribute to the future management of prostate cancer.

ABSTRACT

The global burden of cancer continues to increase, largely because of the aging and growth of the world population. An estimated 14.1 million new cases of cancer occurred worldwide in 2012 [1]. Prostate cancer is the 2nd most common cancer among men and constitutes a substantial health care problem. Given the huge number of affected individuals, current challenges in prostate cancer include imaging methods for improved diagnosis and risk stratification, assessment of response to therapy, and identification of early recurrence. Magnetic resonance imaging (MRI) has become a cornerstone in all phases of cancer imaging due to its versatility and excellent soft tissue contrast. MRI provides anatomical imaging, but also assessment of multiple physiological parameters such as vascularization, cellularity and metabolism [2, 3] by exogenous and endogenous contrast mechanisms. Recommended use of MRI in prostate cancer consists of multi-parametric MRI (mp-MRI) which includes anatomical high-resolution T2-weighted images (T2WI) with functional information from diffusion weighted images (DWI) and/or dynamic contrast enhanced (DCE) MRI [4]. mp-MRI is increasingly being used in the clinical management of prostate cancer for detection and localization, but also to assess the stage and aggressiveness of the disease. The Prostate Imaging – Reporting and Data System (PIRADS) has been designed to promote standardization and minimize variation in the acquisition, interpretation and reading of mp-MRI [5]. However, detection rates of clinically significant disease still vary considerably among reported studies [6]. Magnetic resonance spectroscopic imaging (MRSI) enables the non-invasive assessment of specific metabolites such as choline, citrate, creatine, and polyamines (spermine, spermidine, and putrecine) in the prostate gland. Prostate cancer usually expresses increased concentration of choline and reduced concentrations of citrate and polyamines. Several studies have shown the benefit of adding this metabolic information to MRI in the evaluation of prostate cancer [7-9] . Further, the number of polyamine-free voxels from MRSI is reported to be positively associated with biochemical recurrence following radical prostatectomy [10]. This is also supported by recent findings from ex vivo MR spectroscopy, showing that the spermine concentration in prostatectomy specimens was an independent prognostic marker of recurrence. The metabolic information may thus additionally offer predictive value to establish preoperative risk assessment nomograms. MRSI was included in the original PIRADS recommendation, but not in the most recent version [5], mainly due to the low practicality of current MRSI methods in routine clinical use. However, the technology of MRSI is constantly developing with improvements in hardware, acquisition and processing methods [11-14], and may still show a vital potential in future management of prostate cancer. Positron emission tomography (PET) has a rapidly evolving role in the assessment of prostate cancer [15]. Recent technological advances allows for simultaneous acquisition of MR and PET data, where a benefit arises from combining the excellent soft tissue contrast of MRI with the high molecular specificity of the PET imaging radiotracer. The role of this hybrid modality has a growing interest particularly in biochemical relapse, but also for the detection and staging of primary prostate cancer [16]. The most widely available PET radiotracer, 18F-fluorodeoxyglucose (FDG), has known limitations in prostate cancer, due to the lack of consistent FDG uptake in the slowly proliferating prostate cancer cells. However, several other radiotracers with uptake mechanisms linked to lipid metabolism, amino acid transport and prostate specific membrane antigen expression has shown promising results. Such molecular imaging techniques may also provide important information for targeted radionuclide therapies. The clinical interpretation of MR and molecular images still relies on the manual reading and qualitative evaluation by experienced radiologists, which is a limited and cost-intensive resource. This process may also underuse the quantitative and multi-parametric nature of the data. By radiomics approaches, large amounts of quantitative features from the images are extracted and further exploited [17]. Such features when used with appropriate machine learning algorithms have the potential to uncover disease characteristics from multi-parametric and multi-modal images that are impossible to observe by the naked eye. In the era of Big Data and artificial intelligence, a paradigm shift in radiology is within reach, with an indispensable potential for directing personalized diagnostics and therapy.

Acknowledgements

Funding from The Norwegian Cancer Society (grant number 100792-2013)