Noninvasive assessment of single kidney glomerular filtration rate using multiple diffusion weighted imaging models

Area of Science:

• Radiology
• Nephrology
• Medical Imaging

Background:

• Assessing single kidney glomerular filtration rate (GFR) is crucial for managing kidney diseases.
• Current methods for GFR assessment can be invasive or lack precision.
• Diffusion weighted imaging (DWI) offers a potential non-invasive approach to evaluate renal function.

Purpose of the Study:

• To evaluate the efficacy of various diffusion weighted imaging (DWI) models in assessing single kidney glomerular filtration rate (GFR).
• To compare the performance of different DWI models and regression algorithms for predicting GFR.
• To explore the potential of DWI as an imaging biomarker for kidney function.

Main Methods:

• Adult patients with kidney diseases underwent 3.0-T MRI with 13 b-value DWI.
• Diffusion parameters were calculated using monoexponential model (MEM), diffusion kurtosis imaging (DKI), stretched exponential model (SEM), and intravoxel incoherent motion (IVIM).
• Split GFRs were measured using 99mTc-DTPA scintigraphy (Gates' method), and regression models (linear, regression tree, Gaussian, SVM) were used for prediction, with leave-one-out cross-validation.

Main Results:

• The intravoxel incoherent motion (IVIM) model combined with support vector machine (SVM) regression demonstrated the best performance (RMSE=0.184, R=0.789).
• Integrating parameters from all four DWI models with SVM regression yielded the highest accuracy (RMSE=0.171, R=0.815).
• These findings highlight the predictive power of DWI parameters for GFR estimation.

Conclusions:

• Diffusion weighted imaging (DWI) characteristics can serve as effective imaging biomarkers for single kidney function assessment.
• The integration of advanced DWI models and machine learning algorithms, like SVM, enhances non-invasive GFR evaluation.
• This approach holds potential for advancing non-invasive diagnostic methodologies in nephrology.