Radiomic-enhanced multimodal ultrasound for early detection of acute kidney injury secondary to renal vein thrombosis: a preclinical diagnostic modeling study

Abstract

Acute kidney injury (AKI) resulting from acute renal vein thrombosis (ARVT) is uncommon, yet it can progress swiftly, requiring prompt diagnosis and intervention. This study aimed to investigate the various multimodal ultrasound techniques, specifically conventional ultrasound (CUS), microvascular flow imaging (MFI), contrast-enhanced ultrasound (CEUS), and shear wave elastography (SWE), in conjunction with radiomics for early diagnosis and assessment of AKI resulting from ARVT using a rabbit model. Twenty healthy adult New Zealand white rabbits with 40 kidneys were included in this study. The left kidneys were designated as the experimental group (n = 20), whereas the right kidneys served as the control group(n = 20). Throughout the study, multimodal ultrasound techniques were employed for image acquisition and analysis. The ultrasound images underwent processing, segmentation, feature extraction, and model construction. The dataset was randomly divided in a 7:3 ratio, and the performance of models was assessed through the Receiver Operating Characteristic Curve (ROC) analysis along with key performance metrics. In CUS images, the experimental group showed notable increases in renal volume, cortical thickness, and enhanced cortical echogenicity (p < 0.001, p = 0.032, p < 0.001). In the CDFI, MFI, and CEUS, the experimental group exhibited significant reductions in blood flow perfusion (p < 0.001). In SWE, Young's modulus values for the cortex, medulla, and sinus were significantly elevated in the experimental group (p < 0.001). The strongest correlations were found for creatinine, renal volume, peak systolic velocity of the arcuate artery, time from peak to half-value of sinus, and Young's modulus value for cortex minimum, with Area Under the Curve(AUC) values of 0.600, 0.868, 0.560, 0.503, and 0.982, respectively. The CUS, CDFI, MFI, CEUS, SWE, and CUS+CDFI+MFI+CEUS+SWE radiomics models demonstrated stronger performance, achieving AUC values of 0.899, 0.861, 0.899, 0.833, 0.861, and 0.734, respectively. Multimodal ultrasound combined with radiomics can significantly improve early diagnosis of AKI following ARVT, providing valuable insights for clinical research.

Keywords: Multimodal; acute kidney injury; acute renal vein thrombosis; radiomics; ultrasound.

Graphical abstract

Figure 1.

Flowchart of the overall experimental process. We first established an experimental rabbit model of acute renal vein thrombosis was established. Second, multimodal ultrasound images were obtained. Subsequently, the animal specimens were subjected to pathological analysis. Finally, the images were converted, segmented for the region of interest, followed by feature extraction and selection, the establishment of radiomic model, and performance evaluation.

Figure 2.

ROC Curves for CUS, CDFI, MFI, CEUS, SWE and CUS+CDFI+MFI+CEUS+SWE radiomic model in the testing cohort.