Diegif: An efficient and secured DICOM to EGIF conversion framework for confidentiality in machine learning training

Abstract

Medical imaging plays a critical role in contemporary healthcare, although it confronts issues relating to storage, security, and confidentiality in machine learning-based diagnostic systems. The proposed framework, Diegif, presents an efficient and safe mechanism for converting DICOM (Digital Imaging and Communications in Medicine) data into EGIF (Encrypted Graphics Interchange Format) files to overcome these challenges. The framework comprises four key components: (1) converting DICOM files to GIF format with encryption, (2) decrypting EGIF files for processing, (3) enabling confidentiality-preserving machine learning training using EGIF data, and (4) facilitating physician diagnosis and report generation based on trained machine learning models. The Diegif framework aims to enhance storage efficiency by decreasing file sizes by 66.32%, thereby improving data transport efficacy and cloud storage affordability while preserving strong encryption for data confidentiality. Pseudocode algorithms are provided for each phase, ensuring reproducibility and transparency. This paper illustrates the framework’s potential to medical image processing, secure storage, and AI-driven diagnostic functions in healthcare.