Real-time abnormal cell detection using a deformable snake model

Abstract

Abnormal cell detection by optical microscopy is widely used, and a large number of biopsies required where the process of determining cell’s position in the histological sample image are very time consuming. In this paper, a proof-of-concept study was done to segmented abnormal cells in real time with high performance metrics focusing on minimizing cost and maximizing efficiency. We developed an improved snake/active contour method for fast cancer cells detection by integrated Canny approach. The implementation of this algorithm on Field-Programmable Gate Array (FPGA) technology substantially decreases the required time for identifying abnormal cells. It also allows an efficient and fast computation of active contour in high throughput image analysis applications where time performance is critical. In order to demonstrate the feasibility of this approach, we implemented the architecture on Xilinx Virtex-FPGAs technology which offers for a scalable and a totally embedded on Chip FPGA. The proposed architecture was validated using 30 images from each histopathological type of colon cells, namely, Benign Hyperplasia (BH), Intraepithelial Neoplasia (IN) and Carcinoma (Ca). This method successfully detected the different cell types with computation time in the order of milliseconds. It was compared with manual cells segmented by using the performance metrics of similarity. The experimental results showed that the proposed snake implementation on FPGA produced accurate and stable results. It was able to rapidly identify multiple cellular types from optical microscope images, and effectively addressed difficult problems such as irregular shapes in carcinoma cells. The short computation time in this method makes it applicable to real-time cancer cell detection.