Parallel approach for tubule grading in breast cancer lesions

Abstract

Sequential algorithms and conventional architectures for grading tubules in digitized images of microscopic slide information of breast cancer malignancies are time-consuming and labor-intensive because of high computational complexity. This paper presents a parallel algorithm and its corresponding VLSI architecture for grading tubules in breast cancer lesions in real time. The algorithm relies on a combination of criteria used by physicians including brightness contrast, dark edge, and uniform density of tubule areas. The algorithm exploits concurrencies in both the space domain and the time domain. It only requires each processing element (PE) to perform three simple operations such as addition, subtraction, and comparison, which makes every PE simple. The essential parallelism in the proposed algorithm facilitates its implementation using VLSI architecture, so that the time complexity is only O(N²/k) if using k × k processing elements, where N is the dimension of the image plane. If k = N, the time complexity will be O(N), compared with O(N²) using a uniprocessor. A series of experiments and simulations have verified the high performance of the proposed algorithm and the parallel VLSI architecture.