Chronic kidney disease (CKD) is considered by the World Health Organization to be one of the most widespread public health challenges. In France, it affects nearly 82,000 people, and the number of patients is increasing by 2% each year. Accurate assessment of renal function and structure remains essential for the diagnosis and prognosis of CKD. Renal function has traditionally been assessed either by estimating the glomerular filtration rate (GFR), based on serum creatinine levels, or through invasive biopsy. On the other hand, advanced imaging, including magnetic resonance imaging (MRI), ultrasound elastography (UE), computed tomography (CT), and scintigraphy (PET, SPECT), offers the possibility of non-invasively obtaining structural, functional, and molecular information that could detect changes in the properties and functionality of renal tissues. In addition to CKD, another kidney disease is kidney cancer. Kidney cancer accounts for approximately 2–3% of all cancers, with a 2% annual increase in incidence. Recently, four-phase computed tomography (CT) has become the standard approach for evaluating renal tumors. Renal cell carcinoma, the most common malignant kidney tumor, accounts for 90% of kidney cancers. Although 13–16% of resected masses are benign, solid renal masses—with the exception of benign, fat-rich angiomyolipomas—are generally considered malignant, requiring surgical removal. Thus, appropriate subtyping of renal tumors is of clinical importance because it affects treatment outcomes and helps avoid unnecessary surgical interventions. Furthermore, the nuclear grade of cancerous tumors is reported as an important factor contributing to prognosis, as accurate preoperative assessment of their aggressiveness can help select the optimal treatment method. Currently, the ability of artificial intelligence (AI) to transform conventional medical imaging into a fully automated diagnostic tool is being extensively studied. In addition to qualitative analysis of renal medical imaging, texture analysis has been integrated into machine learning techniques as a means of quantifying renal tissue heterogeneity, providing a promising complementary tool for predicting renal function decline and identifying renal tumors. The current work aims to evaluate the use of texture analysis integrated into machine learning predictive models in three different applications: the detection of renal dysfunction on functional MRI images, the identification of renal tumor type, and the prediction of nuclear grade of renal cancer on CT images. Texture-based predictive models have shown promise in the non-invasive assessment of renal function on MRI images (accuracy, sensitivity, and AUC up to 98%, 98%, and 1, respectively). They demonstrated excellent performance in distinguishing malignant from benign masses (accuracy, sensitivity, and AUC up to 92%, 97%, and 0.9, respectively) and in the subtyping of malignant and benign renal lesions (accuracy, sensitivity, and AUC up to 95%, 95%, and 1, respectively) on CT slices. They were also able to effectively differentiate renal cancer tumor grades (accuracy, sensitivity, and AUC up to 100%, 100%, and 1, respectively). The proposed framework, which is based on textures and machine learning models, has shown promising results in the detection of CKD, the subtyping of renal lesions, and the identification of nuclear grade. It will assist radiologists in real-world clinical settings by providing them with a decision-support tool.
Author : Israa Alnazer
Management team :
- T. Urruty
- M. Naudin
- C. Fernandez
- P. Bourdon
- C.Guillevin
Thesis defended on December 6, 2021
Accepted publications:
- Israa Alnazer, Pascal Bourdon; Thierry Urruty; Omar Falou; Mohamad Khalil; Ahmad Shahin; Christine Fenandez-Maloigne, « Recent Advances in Medical Image Processing for the Evaluation of Chronic Kidney Disease« , Medical Image Analysis (IF 11.148) Pub Date : 2021-01-09, DOI: https://doi.org/10.1016/j.media.2021.101960
- I. Alnazer, P. Bourdon, T. Urruty, O. Falou, M. Khalil, Ahmad Shahin and Christine Fernandez-
Maloigne, « Renal Function Evaluation Survey: Artificial Intelligence for the Next Decade« , MIBEC 2020, Kuala Lumpur Malaysia, 18 octobre 2020 - I. Alnazer, P. Bourdon, T. Urruty, C. Guillevin, M. Naudin, M. Khalil, A. M. Shahin, O. Falou, Christine Fernandez-Maloigne, « ADC Maps Texture Analysis for the Evaluation of Kidney Function: A Preliminary Study » , IEEE ICABME, Tripoli-Lebanon , October 17-18, 2019