Session Item

Radiomics can provide quantitative features from medical imaging that can be correlated to clinical endpoints. The challenges relevant to the robustness of radiomics features have been analyzed by many researchers, as it seems to be influenced by acquisition and reconstruction protocols, as well as by the segmentation of the region of interest (ROI). Prostate cancer (PCa) represents a difficult playground for this technique, due to the discrepancies in the identification of the cancer lesion and the various acquisition protocols. The aim of this study is to investigate the reliability of radiomics in prostate cancer detection.

A homogeneous cohort of patients, with a prostate-specific antigen (PSA) rise that underwent multiparametric MRI imaging of the prostate before prostate biopsy, was tested in this study. All the patients were acquired with the same MRI scanner, with a standardized protocol. The identification of an MRI suspicious cancer lesion was done by two Radiologists with great experience in prostate cancer (>10 years). The segmentation of the lesion was done by two Residents (in Radiation Oncology and Radiologist). After the segmentation, texture features were extracted with LIFEx software. All the patients underwent random prostate biopsy procedures and the presence of prostate cancer, as well as the Gleason score, was retrospectively collected. Texture features were then tested with intraclass coefficient correlation (ICC) analysis to analyze the reliability of the segmentation. 

Forty-four consecutive patients with suspect PCa were included in the present analysis. In 26 patients (59,1%) the prostate biopsy confirmed the presence of PCa, which was scored as Gleason 6 in 6 patients (13,6%), Gleason 3+4 in 8 patients (18,2), and Gleason 4+3 in 12 patients (27.3%). The blind analysis conducted by two physicians as well as the ICC distribution led us to consider the ADC and the DWI400 sequences as the most reliable sequences. There were significant differences in the distribution of ICC in GLCM features (p:0,018), with no differences in the other subsections of Shape features (p:0,893), GLRLM (p:0,594), NGLDM (p:0,109), GLZLM (p:0,594). The reliability analysis, conversely, showed poor results in the majority of the other MRI acquisitions (61% in T2, 89% in DWI50, 44% in DWI400, and 83% in DWI1500), with ADC acquisition only showing better reliability. The ICC distribution of the GLCM features in five different acquisitions is shown in figure 1.

  .

The low ratio of reliability in a monoinstitutional homogenous cohort represents a significant alarm bell for the application of MRI Radiomics in the field of prostate cancer. Supplementary work is needed in a clinical setting to further study the potential of MRI radiomics in PCa.