Deep learning-based Artificial Intelligence algorithms have surpassed human-level performance in many fields including medicine. Specifically in diagnosis using radiology images, deep neural networks empowered AI to excel by educating intricate nonlinear relationships which is a core part of the complicated radiology problems. However, these models require a massive amount of quality data for training. The accuracy of the deep learning model is based on the amount of training data and the quality of the trained data being fed. So, preprocessing the data from different capturing devices is inevitable. This study aimed to highlight some of the image quality metrics that can be used to quantify the efficiency of the chosen preprocessing pipeline. By quantifying the result of each preprocess step, the user can choose an optimal set of preprocesses that can greatly improve the image quality, leading to a high and accurate diagnosis through a deep learning model. Thus, this study detailed how the full reference image quality metrics can be used to validate the performance of sMRI preprocess tasks.
No takes yet. Share an insight, caveat, or question.
Sujana et al. (2024) studied this question.