WISE: A Framework for Gigapixel Whole-Slide-Image Lossless Compression

TL;DR

WSIs pose severe storage challenges due to their gigapixel scale. The paper introduces WISE, a lossless WSI-specific compressor that counteracts information irregularity with a hierarchical projection coding stage, followed by bitmap encoding and a dictionary-based LZW coder. Empirical results across diverse datasets show substantial gains over both traditional and neural compressors, with average improvements around 36× and peaks up to 136×, along with favorable throughput and memory profiles. The work demonstrates practical potential for scalable, lossless WSI storage in real-world digital pathology pipelines.

Abstract

Whole-Slide Images (WSIs) have revolutionized medical analysis by presenting high-resolution images of the whole tissue slide. Despite avoiding the physical storage of the slides, WSIs require considerable data volume, which makes the storage and maintenance of WSI records costly and unsustainable. To this end, this work presents the first investigation of lossless compression of WSI images. Interestingly, we find that most existing compression methods fail to compress the WSI images effectively. Furthermore, our analysis reveals that the failure of existing compressors is mainly due to information irregularity in WSI images. To resolve this issue, we developed a simple yet effective lossless compressor called WISE, specifically designed for WSI images. WISE employs a hierarchical encoding strategy to extract effective bits, reducing the entropy of the image and then adopting a dictionary-based method to handle the irregular frequency patterns. Through extensive experiments, we show that WISE can effectively compress the gigapixel WSI images to 36 times on average and up to 136 times.

Figures (6)

• Figure 1: Illustration of the WISE pipeline. WISE compress the input WSI images through three key steps. As WSI images exhibit a high degree of information imbalance, the first step in WISE is effective information extraction, which performs image preprocessing, hierarchical projection encoding, and bitmap encoding, to transform the input image into information-dense encodings. To be specific, effective bits are gathered together for better pattern extraction. Then, in the second step, WISE constructs a dictionary to capture the byte pattern in transformed image for further compression.
• Figure 2: WSI image's pixel intensity variation is quite different from normal image. It has a large portion of high-frequency features thus affecting the performance of entropy-based compressors.
• Figure 3: Detailed illustration of the Hierarchical Projection Coding, bitmap coding and dictionary coding on a sample matrix. Starting from the original information (left), Hierarchical Projection Coding performs row projection, column projection, and channel projection sequentially, which gradually reduces the entropy of the projected encodings. Then, the bitmap coding transposes the effective bits, coding the matrix with small values to a highly repetitive matrix, as shown in the bottom middle. This approach would increase the byte-based entropy since the repetitive byte is reduced. Nonetheless, after dictionary coding, the entropy would be further reduced.
• Figure 4: Left: Compression ratio variations on different image sizes. Right: WISE's memory usage and throughput on different image sizes.
• Figure 5: PSNR of bitmaps before Bitmap Coding. As illustrated, bitmaps in higher bit positions have larger empty areas and higher PSNR values, indicating they contain less information..