Physics-Informed Deep Learning–Empowered Hybrid IVIM–DKI Using Four Nonzero b-Values for Preoperative Identification of Microvascular Invasion in Hepatocellular Carcinoma

Intravoxel incoherent motion (IVIM) and diffusion kurtosis imaging (DKI) are well-established diffusion-weighted imaging (DWI) techniques, but both typically require multi–b-value acquisitions and computationally expensive parameter estimation. To develop a physics-informed deep learning–empowered hybrid IVIM–DKI framework (PI-DL-IVIM–DKI) for efficient parameter estimation and preoperative identification of microvascular invasion (MVI) in hepatocellular carcinoma (HCC). PI-DL-IVIM–DKI consists of two modules: (1) an optimal b-value design module combining Fisher information matrix analysis and Monte Carlo simulation; and (2) an end-to-end neural network module that estimates hybrid IVIM–DKI parameters based on the optimized b-value set. Parameter fitting and time efficiency were evaluated in both synthetic simulations and in vivo cohort (74 subjects; 56.1±10.6 years), using conventional nonlinear least squares (NLLS) fitting as the reference. The diagnostic performance of PI-DL-IVIM–DKI–derived parameters for preoperative MVI identification was further assessed. The 5-b-value protocol B5 (b = 0, 40, 200, 800, 2000 s/mm 2 ) was selected as the optimal scheme. The deep neural network (DNN) achieved the most accurate parameter estimation under B5 compared with a convolutional network (ConvNet) and a Transformer, and was therefore used as the backbone of PI-DL-IVIM–DKI. PI-DL-IVIM–DKI yielded higher fitting accuracy (NRMSE = 0.156) and substantially improved computational efficiency (up to 1.2×10 5 -fold speed-up) relative to NLLS. In vivo, PI-DL-IVIM–DKI parameters agreed well with the conventional “full-b-value-NLLS” strategy (Pearson r = 0.812–0.993). PI-DL-IVIM–DKI-derived parameters further demonstrated good performance for identifying MVI, with AUC up to 0.765. PI-DL-IVIM–DKI enables accurate and efficient hybrid IVIM–DKI parameter estimation using only four non-zero b-values. PI-DL-IVIM–DKI–derived parameters show potential as imaging biomarkers for MVI in HCC. 1. PI-DL-IVIM–DKI estimates IVIM–DKI parameters using only four non-zero b-values. 2. PI-DL-IVIM–DKI shows high agreement with conventional “full-b-value-NLLS” strategy and provides up to 1.2×10 5 -fold acceleration. 3. PI-DL–IVIM–DKI–derived parameters may facilitate noninvasive preoperative identification of MVI-positive HCC. Summary: A physics-informed deep learning–empowered hybrid IVIM–DKI framework (PI-DL-IVIM–DKI) was developed to achieve accurate and efficient parameter estimation using only four non-zero b-values, and the derived parameters may facilitate noninvasive preoperative identification of MVI in patients with HCC.