Quantitative Analysis With Time-Dependent Diffusion MRI for Assessing WHO/ISUP Tumor Grade in Clear Cell Renal Cell Carcinoma

Abstract

Objective: To investigate the feasibility of time-dependent diffusion-weighted imaging (t_d-dMRI) in assessing the pathological World Health Organization/International Society of Urological Pathology (WHO/ISUP) grade of clear cell renal cell carcinoma (ccRCC).

Materials and methods: A total of 138 patients (median age, 58 years [interquartile range, 51-64 years]; 89 males) with surgically confirmed ccRCC, comprising 48 high-grade (WHO/ISUP grade III/IV) and 90 low-grade (WHO/ISUP grade I/II) tumors, were included in the study among patients who underwent preoperative t_d-dMRI for suspected RCC between May 2022 and May 2024. The t_d-dMRI microstructural parameters, including cell diameter (d), intracellular volume fraction (f_in), cellularity, and extracellular diffusivities (D_ex), were quantified using a two-compartment model. The solid tumor area was manually annotated to extract the mean values from each parameter map. We analyzed the differences in t_d-dMRI parameters between the high- and low-grade tumors and evaluated the ability of these parameters to distinguish between the two tumor groups. High-definition hematoxylin-and-eosin-stained slides were obtained from 92 patients. We assessed the correlation between t_d-dMRI parameters and pathologic nuclear fraction, which was quantified using an automated nucleus segmentation model (Hover-Net).

Results: Compared to high-grade tumors, low-grade tumors exhibited lower cellularity and f_in and higher diameter and D_ex. For differentiation between low- and high-grade ccRCC, the f_in exhibited the highest diagnostic performance (areas under the receiver operating characteristic curve [AUC] = 0.943; 95% confidence interval, 0.906-0.980), followed by cellularity (AUC = 0.931; 0.887-0.976), D_ex (AUC = 0.863; 0.800-0.926), and diameter (AUC = 0.690; 0.596-0.784). The nuclei on pathology slides were automatically segmented, and the nuclear fraction exhibited a moderate correlation with f_in (r = 0.65, P < 0.001).

Conclusion: t_d-dMRI parameters show potential for assessing pathological WHO/ISUP grades and may serve as promising noninvasive biomarkers for characterizing RCC.

Keywords: Diffusion weighted image; Renal cell carcinoma; Time-dependent diffusion; WHO/ISUP grade.

Fig. 1. The inclusion and exclusion flowchart of the participants. RCC = renal cell carcinoma, OGSE = oscillating gradient spin-echo, ccRCC = clear cell renal cell carcinoma, T2WI = T2-weighted imaging

Fig. 2. Schematic illustration of the time-dependent diffusion MRI model fitting process. OGSE = oscillating gradient spin-echo, PGSE = pulsed gradient spin-echo, f_in = intracellular volume fraction, S_in = intracellular diffusion MRI signal, S_ex = extracellular diffusion MRI signal, D_ex = extracellular diffusivity, d = diameter

Fig. 3. The quantitative mappings of T2WI, D_ex, cellularity, d, f_in, D_0Hz, D_17Hz, D_33Hz for patients with WHO/ISUP 1-4. The units for D_ex, D_0Hz, D_17Hz, and D_33Hz were µm²/ms; the d was µm, and cellularity was µm^-1. T2WI = T2-weighted imaging, D_ex = extracellular diffusivity, d = diameter, f_in = intracellular volume fraction, D_0Hz = diffusivity at 0 Hz, D_17Hz = diffusivity at 17 Hz, D_33Hz = diffusivity at 33 Hz, WHO/ISUP = World Health Organization/International Society of Urological Pathology

Fig. 4. The violin boxplots show time-dependent diffusion MRI-based D_ex, cellularity, f_in, d, D_0Hz, D_17Hz, D_33Hz from high- and low-grade clear cell renal cell carcinomas. ^*P < 0.0001, ^†P < 0.01, ^‡P < 0.001, ^§P < 0.05. D_ex = extracellular diffusivity, f_in = intracellular volume fraction, d = diameter, D_0Hz = diffusivity at 0 Hz, D_17Hz = diffusivity at 17 Hz, D_33Hz = diffusivity at 33 Hz, WHO/ISUP = World Health Organization/International Society of Urological Pathology

Fig. 5. The receiver operating characteristic curves of quantitative parameters from time-dependent diffusion MRI in differentiating low- from high-grade clear cell renal cell carcinomas. D_ex = extracellular diffusivity, f_in = intracellular volume fraction

Fig. 6. The workflow of pathology validation. A: A rectangular ROI was placed on the hematoxylin and eosin staining slice to cover the solid area as large as possible. B: The ROI was separated into 2048 × 2048 pixel-sized patches. C: The patches were input into the open-source pre-trained Hover-Net to get the nuclei segmentation. The nuclear fraction on pathology was calculated as the ratio of the sum of the nuclei area to the sum of the tissue area across all patches in 1 slide. D: The correlation map between f_in from the IMPULSED model and the nuclear fraction on pathology. ROI = region of interest, f_in = intracellular volume fraction, IMPULSED = imaging microstructural parameters using limited spectrally edited diffusion