Beyond Multiparametric MRI and towards Radiomics to Detect Prostate Cancer: A Machine Learning Model to Predict Clinically Significant Lesions

Abstract

The risk of misclassifying clinically significant prostate cancer (csPCa) by multiparametric magnetic resonance imaging is consistent, also using the updated PIRADS score and although different definitions of csPCa, patients with Gleason Grade group (GG) ≥ 3 have a significantly worse prognosis. This study aims to develop a machine learning model predicting csPCa (i.e., any GG ≥ 3 lesion at target biopsy) by mpMRI radiomic features and analyzing similarities between GG groups. One hundred and two patients with 117 PIRADS ≥ 3 lesions at mpMRI underwent target+systematic biopsy, providing histologic diagnosis of PCa, 61 GG < 3 and 56 GG ≥ 3. Features were generated locally from an apparent diffusion coefficient and selected, using the LASSO method and Wilcoxon rank-sum test (p < 0.001), to achieve only four features. After data augmentation, the features were exploited to train a support vector machine classifier, subsequently validated on a test set. To assess the results, Kruskal–Wallis and Wilcoxon rank-sum tests (p < 0.001) and receiver operating characteristic (ROC)-related metrics were used. GG1 and GG2 were equivalent (p = 0.26), whilst clear separations between either GG[1,2] and GG ≥ 3 exist (p < 10−6). On the test set, the area under the curve = 0.88 (95% CI, 0.68–0.94), with positive and negative predictive values being 84%. The features retain a histological interpretation. Our model hints at GG2 being much more similar to GG1 than GG ≥ 3.

An Apparent Diffusion Coefficient-Based Machine Learning Model Can Improve Prostate Cancer Detection in the Grey Area of the Prostate Imaging Reporting and Data System Category 3: A Single-Centre Experience

“…Nevertheless, the polynomial model should be preferred in order to have a balanced distribution of FP and FN errors, which yields more similar values of SP and SN, and of PPV and NPV values, accordingly. Our findings also confirm the well-established role of ADC radiomics, already exploited in a previous study [ 33 ], and highlight the feasibility of single-sequence radiomics, thus also minimizing, in the view of a prospective implementation of radiomic tools in the clinical practice, the computing time, favouring a reduction in patient motion, and the time required to clinicians for image series segmentation, after acquisition.…”

“…Quantitative imaging features, so-called radiomic features (RFs), were generated according to our method proposed in [ 30 ], already applied in [ 31 , 32 , 33 ]. The method is based on a two-stage procedure.…”

“…First, 12 first-order features (i.e., mean, median, entropy, uniformity, standard deviation, median absolute deviation, interquartile range, coefficient of variation, skewness, kurtosis, mean and median of the last decile of image pixel distribution) were computed locally within each prostate ROI, that is, by assigning each ROI pixel the value of a first-order metric computed in a rectangular window (centred on the pixel itself). In particular, a window of 9 × 9 pixel size was adopted, based on a previous report [ 33 ]. This stage provided us with 12 stacks of parametric maps for each patient, displayed as hot–cold colorimetric maps, and as many local first-order features, for all prostate and lesion slices.…”

Development trends and knowledge framework in the application of magnetic resonance imaging in prostate cancer: a bibliometric analysis from 1984 to 2022

“…The ML technique is a representative method for exploring the risk factors or high-risk groups of disease by analyzing medical big data ( 52 ). It often combines with radiomics, and plays an important role in the accurate diagnosis and treatment of PCa by building various models ( 53 - 55 ). DL is an important branch of AI that uses networks of simple interconnected units to extract patterns from data to solve complex problems ( 56 ).…”

Radiomic Features from Post-Operative 18F-FDG PET/CT and CT Imaging Associated with Locally Recurrent Rectal Cancer: Preliminary Findings

“…The application of well-established machine learning and artificial intelligence techniques to medical image analysis, nowadays known as radiomics, notably enriches the information retrievable from different types of clinical images (e.g., CT, MR, and PET images) and ultimately improves the diagnostic potential of the imaging modalities. In fact, the Radiomic Features (RFs) extracted from routinely acquired medical images enable a quantitative and objective characterization of tissue properties, latent ones included [11]. Accordingly, RFs become potential promotors of predictive imaging biomarkers, thereby allowing the early detection of LRRC.…”