Full Field Digital Mammography Authentication Method Based on Autoencoder and Principal Component Extraction
Binjia Zhou, Qirui Wang, Qianwei Zhou · 2024
Electronic Health Record (EHR) systems have replaced paper-based systems in healthcare organizations. There are now many measures in the medical field to protect sensitive big data. However, the healthcare industry is still the most affected by privacy breaches, and manipulations of fake data are endless. Content-Based Image Retrieval (CBIR) enables the retrieval of images similar to the query image and facilitates the comparison of current cases with historical cases stored in the database. This technique can find valuable applications in medical image verification, particularly in tackling the issue of counterfeit images. For image validation, we apply autoencoder to extract features, principal component analysis and genetic algorithm to optimize the extracted features, and then hand the features to vision transformer to make decisions. As a result, in comparative experiments, it is proved that our method conforms to the standard and has good performance.