Publication

Abstract : Computed Tomography (CT) Perfusion is an advanced form of Contrast-Enhanced CT imaging for evaluating the cerebral blood flow dynamics during the assessment of brain strokes. However, when compared to a plain Non-Contrast CT (NCCT) scan which is cheaper and faster, such Contrast-Enhanced CT (CECT) scans pose significant risks for vulnerable patient groups, despite the precision they offer. In this work, we leverage a deep transfer learning method for the automated generation of Contrast-Enhanced CT (CECT) images from Non-Contrast CT (NCCT) scans, effectively bypassing the costs, risks, and variable time delays in performing a Contrast-Enhanced CT scan. We compare the performance of different state-of-the-art ImageNet models for generating CT perfusion maps from NCCT images. Threefold cross-validation is employed to assess the model's performance on unseen data, with Mean Squared Error (MSE), Structural Similarity Index Measure (SSIM), and Peak Signal-to-Noise Ratio (PSNR) as the evaluation metrics. The model utilizing the ResNet-50 architecture effectively maps and delineates CECT images from Non-Contrast CT (NCCT) scans, demonstrating a strong performance with 26.4dB PSNR and 89.49% SSIM respectively.