Acceleration of Image Segmentation Algorithm for (Breast) Mammogram Images Using High-Performance Reconfigurable...

Area of Science:

• Medical Imaging
• Computer Engineering
• Biomedical Signal Processing

Background:

• Image segmentation is crucial for medical imaging, particularly in breast cancer detection from mammograms.
• Mammography analysis involves multiple stages, including region extraction, mass identification, and classification, which can be time-consuming with large datasets.

Purpose of the Study:

• To implement and evaluate an existing region-of-interest based image segmentation algorithm for mammograms on High-Performance Reconfigurable Dataflow Computers (HPRDCs).
• To investigate the acceleration achieved by utilizing Maxeler's acceleration card as a dataflow engine (DFE) within the HPRDC framework.

Main Methods:

• The study involved implementing a region-of-interest image segmentation algorithm on HPRDCs.
• Maxeler's acceleration card was employed as the dataflow engine (DFE).
• Experiments were conducted using mammogram images of varying resolutions, with different DFE configurations tested.

Main Results:

• Experimental results demonstrated significant acceleration of the image segmentation algorithm when executed on HPRDCs.
• Different DFE configurations yielded varying degrees of acceleration, indicating the impact of hardware setup on performance.
• The implemented algorithm showed good acceleration, highlighting the potential of HPRDCs for medical imaging tasks.

Conclusions:

• HPRDCs, utilizing Maxeler's acceleration cards, offer a viable solution for accelerating computationally intensive tasks in medical image analysis.
• The acceleration achieved in mammogram segmentation can contribute to faster and more efficient breast cancer detection workflows.
• Further optimization and exploration of DFE configurations can potentially yield even greater performance improvements.