Accelerating Pathology Image Data Cross-Comparison on CPU-GPU Hybrid Systems.

Area of Science:

• Computational pathology
• Bioinformatics
• Database systems

Background:

• Spatial databases are crucial for pathology imaging analysis.
• Current systems struggle with the data and compute demands of cross-comparing micro-anatomic objects.
• Existing implementations do not fully leverage modern parallel hardware.

Purpose of the Study:

• To develop a cost-effective software solution to accelerate spatial cross-comparison operations.
• To enhance the performance of spatial database systems in pathology imaging analysis.
• To exploit Graphics Processing Units (GPUs) and multi-core Central Processing Units (CPUs) for improved throughput.

Main Methods:

• Developed an efficient Graphics Processing Unit (GPU) algorithm.
• Implemented a pipelined system framework with task migration support.
• Utilized a combination of GPUs and multi-core CPUs for parallel processing.

Main Results:

• The proposed solution significantly accelerates spatial cross-comparison.
• Performance improvements exceed 18 times compared to a parallelized spatial database approach.
• Demonstrated effectiveness using extensive experiments with real-world pathology data.

Conclusions:

• The customized software solution offers a cost-effective method for accelerating spatial cross-comparison.
• The approach effectively utilizes modern parallel hardware for high-throughput analysis.
• This work addresses the performance limitations of current spatial database systems in pathology imaging.