ABSTRACT
In Europe, prostate cancer (PCa) is the most common solid neoplasm, with an incidence rate of 214 cases per 1000 men, outnumbering lung and colorectal cancer. PCa affects elderly men more often and therefore is a bigger health concern in developed countries (Heidenreich et al. 2011). PCa diagnostics are initiated on the basis of prostatespecific antigen (PSA) measurements and determination of clinical stage by means of digital rectal examination (DRE). Definite diagnosis is usually obtained by means of transrectal ultrasonography (TRUS)-guided systematic random prostate biopsies. Histopathologic analysis of these biopsy samples provides the clinician with information on the Gleason score (GS), a histopathologic score that correlates with biologic activity and aggressiveness. According to current international convention, the GS of cancers detected in a prostate biopsy consists of the Gleason grade of the dominant (most extensive) carcinoma component plus the highest grade among the remaining patterns, regardless of its extent. Nomograms based on the combination of PSA level, DRE findings, and biopsy GS are
and heterogeneity of PCa. For this reason, Partin tables and risk stratification schemes that incorporate information from biopsy-determined Gleason grades into decision making are rendered less accurate and less reliable. There is a definite need for a method with which to improve the accuracy in determining the risk of progression before treatment (Hambrock et al. 2011). Currently, if clinical suspicion for PCa persists in spite of negative prostate biopsies, a patient may undergoes Magnetic Resonance Imaging (MRI) examination, using a multiparametric (mp) approach, which combines anatomical and functional data. A second biopsy is then performed exploiting the information on tumour localization provided by the mp-MRI. The advent of mp-MRI suggests an increased role for imaging also in risk stratification and treatment planning (Turkbey et al. 2009). Specifically, recentstudies have shown that quantitative MRI metrics may serve as non-invasive biomarkers of tumor aggressiveness, with the potential to complement biopsy and PSA findings in guiding management (Hambrock et al. 2011, Rosenkrantz et al. 2012). However, these preliminary results have been assessed using TRUS biopsy as reference standard, which is known to result in undergrading in a fraction of tumors in comparison with prostatectomy (Rosenkrantz et al. 2012). Considering T2-weighted (T2w) images, it is well know that PCa usually shows a lower signal intensity (SI) than non-neoplastic prostatic tissue, while only one study investigated retrospectively whether the SI of PCa correlates with the Gleason grade at wholemount step-section pathologic evaluation after RP (Wang et al. 2008). The purpose of this study is to differentiate PCa aggressiveness from T2w MRI in a dataset including also intermediate-risk PCa, exploiting an algorithm pipeline completely automatic and therefore easily integrable in Computer Aided Diagnosis (CAD) systems and in the clinical routine practices.
