Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Efficiency vs. Efficacy: Assessing the Compression Ratio-Dice Score Relationship through a Simple Benchmarking Framework for Cerebrovascular 3D Segmentation

Shimaa Elbana, Ahmad Kamal, Shahd Ahmed Ali, Ahmad Al-Kabbany

TL;DR

This study assesses how ZFP compression affects automated 3D cerebrovascular segmentation on a large RSNA dataset. By comparing fixed-rate and error-tolerance modes using two Mamba-based segmentation architectures, it quantifies the trade-offs between compression ratio and task fidelity via Dice and IoU metrics. The findings show that error-tolerance mode achieves up to ~49:1 compression with only minor Dice reductions, while fixed-rate compression remains robust up to ~8:1, indicating strong task robustness to lossy compression. The work demonstrates a practical, training-free approach to data reduction that can accelerate collaboration and data sharing in large-scale medical imaging research.

Abstract

The increasing size and complexity of medical imaging datasets, particularly in 3D formats, present significant barriers to collaborative research and transferability. This study investigates whether the ZFP compression technique can mitigate these challenges without compromising the performance of automated cerebrovascular segmentation, a critical first step in intracranial aneurysm detection. We apply ZFP in both its error tolerance and fixed-rate modes to a large scale, and one of the most recent, datasets in the literature, 3D medical dataset containing ground-truth vascular segmentations. The segmentation quality on the compressed volumes is rigorously compared to the uncompressed baseline (Dice approximately equals 0.8774). Our findings reveal that ZFP can achieve substantial data reduction--up to a 22.89:1 ratio in error tolerance mode--while maintaining a high degree of fidelity, with the mean Dice coefficient remaining high at 0.87656. These results demonstrate that ZFP is a viable and powerful tool for enabling more efficient and accessible research on large-scale medical datasets, fostering broader collaboration across the community.

Efficiency vs. Efficacy: Assessing the Compression Ratio-Dice Score Relationship through a Simple Benchmarking Framework for Cerebrovascular 3D Segmentation

TL;DR

This study assesses how ZFP compression affects automated 3D cerebrovascular segmentation on a large RSNA dataset. By comparing fixed-rate and error-tolerance modes using two Mamba-based segmentation architectures, it quantifies the trade-offs between compression ratio and task fidelity via Dice and IoU metrics. The findings show that error-tolerance mode achieves up to ~49:1 compression with only minor Dice reductions, while fixed-rate compression remains robust up to ~8:1, indicating strong task robustness to lossy compression. The work demonstrates a practical, training-free approach to data reduction that can accelerate collaboration and data sharing in large-scale medical imaging research.

Abstract

The increasing size and complexity of medical imaging datasets, particularly in 3D formats, present significant barriers to collaborative research and transferability. This study investigates whether the ZFP compression technique can mitigate these challenges without compromising the performance of automated cerebrovascular segmentation, a critical first step in intracranial aneurysm detection. We apply ZFP in both its error tolerance and fixed-rate modes to a large scale, and one of the most recent, datasets in the literature, 3D medical dataset containing ground-truth vascular segmentations. The segmentation quality on the compressed volumes is rigorously compared to the uncompressed baseline (Dice approximately equals 0.8774). Our findings reveal that ZFP can achieve substantial data reduction--up to a 22.89:1 ratio in error tolerance mode--while maintaining a high degree of fidelity, with the mean Dice coefficient remaining high at 0.87656. These results demonstrate that ZFP is a viable and powerful tool for enabling more efficient and accessible research on large-scale medical datasets, fostering broader collaboration across the community.

Paper Structure

This paper contains 9 sections, 4 figures, 4 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Materials and Methods
  4. Dataset
  5. Compression Methodology
  6. Segmentation and Evaluation
  7. Implementation and Training Protocols
  8. Experimental Results
  9. Conclusion

Figures (4)

  • Figure 1: Workflow of the proposed framework for vessel segmentation under compression.
  • Figure 2: The dual-axis plot displays the relationship between the Compression Ratio (CR, Left Y-axis) and the Mean DICE Score (Right Y-axis) as the compression rate (bits/voxel) is aggressively reduced. While the CR exhibits a steep, near 8-fold increase, the Mean DICE Score remains remarkably stable and visually flat across all tested rates, demonstrating that the critical features for segmentation are preserved despite aggressive lossy compression.
  • Figure 3: This dual-axis chart compares the exponential increase in the Compression Ratio (CR, Left Y-axis) against the decline in the Mean DICE Score (Right Y-axis) as the absolute error tolerance (Ts) is increased. The CR rises dramatically, reaching nearly 50:1 at Ts=1500, while the Mean DICE Score experiences a controlled and relatively moderate drop, confirming that high-fidelity segmentation can be maintained even under extreme error-bounded compression.
  • Figure 4: Inference results of VesselMamba++ on MRA data under different compression settings. (a) Example 1 and (b) Example 2 show comparisons between the original vessel mask, inference without compression, absolute error tolerance levels (500, 1000, 1500), and fixed error rates (4, 8, 16). Dice scores are displayed above each prediction, demonstrating that vessel structure is well preserved even under lossy compression.