Coronary artery properties in atherosclerosis: A deep learning predictive model

FIGURE 1. Idealized atherosclerotic coronary artery including lipid-rich core, fibrotic tissue, fibrous cap, tunica media, and tunica adventitia.

FIGURE 2. The workflow is divided into building the database, generating the ANN, and applying it to realistic geometries: 1) Building the idealized database and performing the FEA simulations; 2) Selecting the outputs of the FEA (ϵ 1, ϵ 2, ϵ core , ϵ plaque , ϵ theta , and SR); 3) Statistical analysis of the FEA output variables and selection of the inputs of the ANN ( ϵcore2 , ϵplaque2 , ϵtheta2 , and SR); 4) Selecting MLP structure; 5) Training and validating the ANN; 6) Testing the ANN by computing the relative error in the prediction of E core and E plaque ; 7) Building the realistic database; 8) Application of the ANN to the realistic database; 9) Application of transfer-learning technique; 10) Application of the transfer-learned ANN to the realistic database and computation of the relative errors.

FIGURE 3. Construction of realistic database using 20 cases with varied fibrous cap thickness, based on a real geometry (65μm–300 μm).

FIGURE 4. Descriptive statistical analysis of the candidate variables against the response variables. (A) Maximum Principal Strain ϵ 1 vs. Young’s modulus of the lipid core E core . (B) Minimum Principal Strain ϵ2 vs. Young’s modulus of the lipid core E core . (C) Maximum Principal Strain ϵ1 vs. Young’s modulus of the lipid core E plaque . (D) Minimum Principal Strain ϵ2 vs. Young’s modulus of the lipid core E plaque .

FIGURE 5. Loss function for training and validation step evaluated with the MSE.

FIGURE 6. Relative error in the prediction of E core (blue) and E plaque (green) in the training and testing steps of the ANN.

FIGURE 7. Relative error in the prediction of E core (blue) and E plaque (green) using the transfer-learned ANN with the realistic database.