Lossy Image Compression in a Preclinical Multimodal Imaging Study

Area of Science:

• Biomedical Imaging
• Data Science
• Pre-clinical Research

Background:

• High-resolution volumetric data in pre-clinical studies present significant storage and handling challenges.
• Lack of standardized guidelines for image compression in pre-clinical research exacerbates data management issues.
• Lossy compression offers a potential solution for alleviating these data-intensive challenges.

Purpose of the Study:

• To investigate the application and impact of lossy image compression on high-resolution volumetric biomedical data.
• To quantify the effects of compression on data metrics and expert interpretation in pre-clinical studies.
• To establish trade-offs between data reduction and preservation of visual information for volumetric datasets.

Main Methods:

• Applied lossy image coding to compress volumetric data from high-resolution episcopic microscopy (HREM), micro-computed tomography (µCT), and micro-magnetic resonance imaging (µMRI).
• Assessed compression impact by measuring task-specific performance of biomedical experts interpreting and labeling compressed data volumes.
• Quantified compression effects using Jaccard Index (JI) and average Hausdorff Distance (HD) after vasculature segmentation.

Main Results:

• Defined trade-offs between data volume reduction and preservation of visual information, ensuring relevant vasculature morphology retention.
• Demonstrated that a 256-fold data size reduction maintained compression-induced error below inter-observer variability.
• Showcased minimal impact on the assessment of murine tumor vasculature across scales despite significant compression.

Conclusions:

• Lossy compression is a viable strategy for managing large pre-clinical volumetric datasets.
• Compression up to a 256-fold reduction can be employed without compromising the accuracy of tumor vasculature assessment.
• This approach balances data volume reduction with the preservation of essential morphological details in pre-clinical imaging.