SearcharxivSearch

arXiv subjects

Bachir Taouli

Publications and source records attributed to Bachir Taouli.

3 recordsLinked to original sources

Detecting Early Kidney Allograft Fibrosis with Multi-b-value Spectral Diffusion MRI

Kidney allograft fibrosis is a marker of chronic kidney disease (CKD) and predicts functional decline, and eventual allograft failure. This study evaluates if spectral diffusion MRI can help detect early development and mild/moderate fibrosis in kidney allografts. In a prospective two-center study of kidney allografts, interstitial fibrosis and tubular atrophy (IFTA) was scored and eGFR was calculated from serum creatinine. Multi-b-value DWI (bvalues=[0,10,30,50,80,120,200,400,800mm2/s]) was post-processed with spectral diffusion, intravoxel incoherent motion (IVIM), and apparent diffusion coefficient (ADC). Connection between imaging parameters and biological processes was measured by Mann-Whitney U-test and Spearman's rank; diagnostic ability was measured by five-fold cross-validation univariate and multi-variate logistic regression. Quality control analyses included volunteer MRI (n=4) and inter-observer analysis (n=19). 99 patients were included (50$\pm$13yo, 64M/35F, 39 IFTA=0, 22 IFTA=2, 20 IFTA=4, 18 IFTA=6, 46 eGFR<=45mL/min/1.73m2, mean eGFR=47.5$\pm$21.3mL/min/1.73m2). Spectral diffusion detected fibrosis (IFTA>0) in patients with normal/stable eGFR>45ml/min/1.73m2 [AUC(95$\%$CI)=0.72(0.56,0.87),p=0.007]. Spectral diffusion detected mild/moderate fibrosis (IFTA=2-4) [AUC(95$\%$CI)=0.65(0.52,0.71),p=0.023], as did ADC [AUC(95$\%$CI)=0.71(0.54,0.87),p=0.013)]. eGFR, time-from-transplant, and allograft size could not. Interobserver correlation was >0.50 in 24 out of 40 diffusion parameters. Spectral diffusion MRI showed detection of mild/moderate fibrosis and fibrosis before decline in function. It is a promising method to detect early development of fibrosis and CKD before progression.

physics.med-ph

Estimation of Multi-Component Flow in the Kidney with Multi-b-value Spectral Diffusion

Purpose: Examine the theory and potential clinical application of estimated intravoxel flow of separated perfusion, tubular flow, and diffusion from multi-b-value DWI in kidney allografts. Methods: Multi-b-value DWI (9 b-values; 0-800 s/mm2) from a kidney cortex is simulated with anisotropic and non-Gaussian (i.e. anomalous) vascular, tubular, and tissue components and analyzed with a Bayesian biexponential, least-squares triexponential, and spectral diffusion MRI. Comparison and application of biexponential, triexponential, and spectral diffusion fD is demonstrated in a two-center study of 54 kidney allografts patients (21F/33M, 48.8 SD 10.5years) and compared to fibrosis (Banff 2017 interstitial fibrosis and tubular atrophy score 0-6 from clinical biopsies of the renal cortex), impaired kidney function (CKD-EPI 2021 eGFR<45ml/min/1.73m2), and proteinuria. Results: Spectral diffusion fD demonstrated strong correlation to input fD of the simulated anisotropic and anomalous components. It agreed with both three-component diffusion and two-component diffusion. fD showed similar or improved agreement and correlation to input compared to individual parameters, and similar or improved agreement to corresponding bi- and triexponential models. In kidney allografts, spectral diffusion fD showed higher allograft fibrosis score had higher fD_tissue, impaired allograft function showed reduced fD_tubule, and fD_vascular negatively correlated with proteinuria across diagnostic groups of function and fibrosis. Conclusions: Spectral diffusion MRI with multi-Gaussian fD as a flow proxy separated different anomalous and anisotropic diffusion components of perfusion, tubular flow, and tissue diffusion and may hold clinical value in diffusion MRI of kidney pathophysiology.

physics.med-ph

MRAnnotator: multi-Anatomy and many-Sequence MRI segmentation of 44 structures

In this retrospective study, we annotated 44 structures on two datasets: an internal dataset of 1,518 MRI sequences from 843 patients at the Mount Sinai Health System, and an external dataset of 397 MRI sequences from 263 patients for benchmarking. The internal dataset trained the nnU-Net model MRAnnotator, which demonstrated strong generalizability on the external dataset. MRAnnotator outperformed existing models such as TotalSegmentator MRI and MRSegmentator on both datasets, achieving an overall average Dice score of 0.878 on the internal dataset and 0.875 on the external set. Model weights are available on GitHub, and the external test set can be shared upon request.

eess.IV