GaborFlow improves brain tumor detection with efficient hardware accelerators
GaborFlow: Hardware-Software Co-Optimization of Gabor Filter Accelerator Design
Hardware Architecture
Summary
Brain tumor detection from MRI scans is important but challenging, especially when accuracy and energy efficiency matter. The authors developed GaborFlow, a new way to design a specialized computer chip and software together to speed up a key step called Gabor filtering. Their system tunes filter settings and uses smart hardware design to improve accuracy and reduce power use. This approach achieved very high tumor classification accuracy while saving energy and space compared to other methods. They share their hardware design openly so others can use and build on it.
What this means in practice
- •For medical device engineers: Design specialized hardware for faster and more accurate brain tumor classification from MRI scans using trainable Gabor filters.
- •For embedded ai developers: Build energy-efficient deep learning accelerators that integrate trainable Gabor filters to enhance medical image feature extraction.
Authors
Priyanka Agarwal, Pruthvi Parate, Raghav Khurana, Madhav Rao
Abstract
Brain tumor classification from MRI scans represents a critical challenge in medical imaging, where accuracy directly impacts patient outcomes. This paper introduces GaborFlow, a novel hardware-software co-optimization framework that revolutionizes the design of the Gabor Filter accelerator for generating feature maps to enhance brain tumor detection. Unlike conventional approaches that rely on fixed filter parameters, our method employs trainable Gabor coefficients within a carefully optimized range, integrated with an EfficientNet-B0 backbone CNN architecture. The innovation lies in the synergistic combination of software optimization through Bayesian hyperparameter tuning and hardware efficiency through precision-optimized multipliers. Our GF baseline accelerator achieves remarkable results, with 98.6% classification accuracy, 28.08% power benefits, and 48.83% footprint savings compared to state-of-the-art implementations, demonstrating that medical AI can be both highly accurate and energy-efficient. The hardware design files are made freely available for easy adoption and further usage by the researchers and designers.