FDIC: An Adaptive End-to-End Compression Framework for Medical Images

This is the authors' abstract. We don't add key points for this paper.

In the era of big data, the demand for the transmission and storage of medical images has surged, posing challenges to the limitations of traditional image compression algorithms.Deep learning-based end-to-end image compression techniques have emerged as a research focus.However, existing end-to-end algorithms lack sufficient adaptability, making it difficult to meet the requirements of complex application scenarios for medical images. This paper focuses on improving the adaptability of image compression algorithms and proposes an efficient image compression algorithm with flexible rate and complexity adjustments, named Flexible Deformable Image Codec (FDIC). The algorithm constructs a basic framework by combining a variable bit rate module with a parameter-sharing mechanism, achieves end-to-end optimization through a multistage training strategy, and demonstrates excellent performance on various medical image datasets. Compared with traditional algorithms and other methods, FDIC can maintain image quality while enabling flexible adjustments of rate and decoding complexity, providing a more efficient solution for the storage and transmission of medical images. This advancement is expected to promote the development of medical imaging informatization.