MRI-Based Grading of Clear Cell Renal Cell Carcinoma Using a Machine Learning Classifier

Abstract

ObjectiveTo develop a machine learning (ML)-based classifier for discriminating between low-grade (ISUP I-II) and high-grade (ISUP III-IV) clear cell renal cell carcinomas (ccRCCs) using MRI textures.Materials and MethodsWe retrospectively evaluated a total of 99 patients (with 61 low-grade and 38 high-grade ccRCCs), who were randomly divided into a training set (n = 70) and a validation set (n = 29). Regions of interest (ROIs) of all tumors were manually drawn three times by a radiologist at the maximum lesion level of the cross-sectional CMP sequence images. The quantitative texture analysis software, MaZda, was used to extract texture features, including histograms, co-occurrence matrixes, run-length matrixes, gradient models, and autoregressive models. Reproducibility of the texture features was assessed with the intra-class correlation coefficient (ICC). Features were chosen based on their importance coefficients in a random forest model, while the multi-layer perceptron algorithm was used to build a classifier on the training set, which was later evaluated with the validation set.ResultsThe ICCs of 257 texture features were equal to or higher than 0.80 (0.828–0.998. Six features, namely Kurtosis, 135dr_RLNonUni, Horzl_GLevNonU, 135dr_GLevNonU, S(4,4)Entropy, and S(0,5)SumEntrp, were chosen to develop the multi-layer perceptron classifier. A three-layer perceptron model, which has 229 nodes in the hidden layer, was trained on the training set. The accuracy of the model was 95.7% with the training set and 86.2% with the validation set. The areas under the receiver operating curves were 0.997 and 0.758 for the training and validation sets, respectively.ConclusionsA machine learning-based grading model was developed that can aid in the clinical diagnosis of clear cell renal cell carcinoma using MRI images.

Role of AI and Radiomic Markers in Early Diagnosis of Renal Cancer and Clinical Outcome Prediction: A Brief Review

“…Chen et al [41] aimed at grading ccRCC using Phase 2 CEMRI. Their study included 99 tumors, with 61 low-grade and 38 high-grade cases.…”

“…This survey reviews the studies from the previous decade that used AI and radiomicbased markers derived from CECT, CEMR, multiparametric MR, or DW-MR imaging modalities to produce AI-based CAD systems for diagnosing RC at an early stage. Specifically, the included studies aimed to identify a given renal tumor malignancy status [14,[27][28][29][30][31][32][33], specify the associated subtype [22,[33][34][35][36], and grade/stage the malignant tumors (I-IV) [22,[37][38][39][40][41]. In addition to accurate diagnosis, treatment follow-up protocol is crucial to evaluate patients' clinical outcome/treatment response, including the recurrence rate, overall survival (OS), and progression-free survival (PFS) rate.…”

“…In both imaging modalities, first-order texture markers, including entropy, mean, MPP, skewness, and kurtosis, were reported to be sufficient for the intended purpose. • For subtyping and grading, both CT [68,71,73,74,76,79,80,[88][89][90][91] and MRI [22,41,95,96] studies exhibited adequate diagnostic performance, suggesting that second-order texture markers, particularly those derived from the GLCM and GLRLM, should be combined with first-order texture markers. A limited number of studies have relied on morphological or functional markers, which, if integrated, could significantly enhance diagnostic performance [68].…”

“…Handcrafted techniques have proven efficient, as evidenced by high diagnostic accuracy, sensitivity, and specificity, as well as being well-understood (i.e., explainable AI), making them desirable and dependable. Suggested Diagnostic Classifiers: The RF, SVM, and ANN classifiers were the most frequently utilized AI-based classification models in CT studies [14,27,28,35,36,38,39,[52][53][54][55]58,59,67,68,70,74,76,[78][79][80]85,86,89,90], while the RF classifier was predominantly selected in MRI studies [29,32,33,41,93,95,96]. These classifiers have provided impressive diagnostic results and have been widely accepted by researchers in the field due to their ability to handle nonlinear and multiclass classification problems.…”

“…These classifiers have provided impressive diagnostic results and have been widely accepted by researchers in the field due to their ability to handle nonlinear and multiclass classification problems. Suggested Imaging Modalities/Phases: Contrast-enhanced phases 2 and 3 (corticomedullary/, arterial phase, and nephrographic/portal venous phase) were reported to be the most informative phases for extracting radiomic markers in both CT [35,36,38,58,62,64,68,[70][71][72][73][74][75][76]80,81,87,[89][90][91] and MRI [41] studies. Meanwhile, texture analysis of ADCs on DW-MRI was the most commonly employed technique to extract radiomic markers in MRI studies [29][30][31][32]94].…”