Advancements in Lossless and Reversible Compression of Digital Pathology Images via Auto-Recursive Set Partitioning in Hierarchical Trees and Wavelet Decomposition

Abstract

Set Partitioning in Hierarchical Trees (SPIHT) represents a leading-edge algorithm in nearlossless image compression, leveraging the Discrete Wavelet Transform. However, its effectiveness diminishes when applied to high-resolution images such as Digital Pathology Images (DPIs). This research aims to enhance the SPIHT algorithm specifically for DPIs by investigating the impact of applying various wavelets in the wavelet decomposition process and the introduction of auto-recursion in the SPIHT algorithm. An extensive selection of wavelet types were tested within the wavelet decomposition process integral to the SPIHT algorithm. The ultimate goal was to identify the wavelet that yields the highest compression ratio and the one that maintains the highest data consistency. The proposed auto-recursion was also examined against the original n-recursive algorithm to discern differences in compression performance. The results indicated that the BIOR 5.5 wavelet is more apt for achieving a high compression ratio, while the BIOR 3.9 wavelet is more suitable for securing high compression quality in the compression of high-resolution DPIs. The newly introduced auto-recursion feature contributes significantly to optimizing the quality of the compressed image. Visual verification of the compressed image's quality, for any potential loss of detail, was carried out through expert validation in a clinical setting. This expert validation confirmed that the proposed algorithm can produce higher quality compressed images with negligible loss of quality. Thus, this research offers a partial solution to current challenges in digital pathology related to storage, transfer, and archiving of high-resolution DPIs, by providing a more effective compression algorithm.