Cancer Cell Detection and Tracking Based on Local Interest Point Detectors

• Type: Article in International Conference Proceedings Book
• Date: 2013

Authors

Esteves, T (Author) Other; Oliveira, MJ (Author) FCUP; Quelhas, P (Author) Other

Conference proceedings International

Title: IMAGE ANALYSIS AND RECOGNITION

Pages: 434-441

10th International Conference on Image Analysis and Recognition (ICIAR)

Pvoa do Varzim, PORTUGAL, JUN 26-28, 2013

Indexing

ISI Web of Knowledge - 2 Citations

Scopus - 1 Citation

Other information

Authenticus ID: P-008-EA6

DOI: 10.1007/978-3-642-39094-4_49

Abstract (EN): The automatic analysis of cell mobility has gained increasing relevance given the enormous amount of data that biology researchers have currently to analyze. However, most biology researchers still analyze cells by visual inspection alone, which is time consuming and prone to induce subjective bias. This makes automatic cell's mobility analysis essential for large scale, objective studies of cells. To evaluate cancer cell's mobility, biologists establish in vitro assays with cancer cells seeded on native surfaces or on surfaces coated with extracellular matrix components, recording time-lapse brightfield microscopy images. In such analysis only through the use of quantitative automatic analysis tools is it possible to gather evidence to firmly support biological findings. In order to perform cell mobility analysis, we perform cell tracking based on cell detection. To detect cells with robustness and increased performance we propose the use of a local interest point detector, the scale-normalized Laplacian of Gaussians filter which enhances the image's blob like structure which corresponds to cell locations. For cells mobility analysis the tracking of cells is performed by a detection association approach assuming either a random or a constant velocity motion and using similarity measures as cross correlation coefficient and SIFT descriptors similarity. Based on experimental results we found that the assumption of a random motion and the use of the SIFT descriptors for the tracking process outperformed all the other approaches obtaining an accuracy in the detection process of 78.6% and considering the tracking, 87.1% of the total number of cell associations between frames were correctly identified.

Language: English

Type (Professor's evaluation): Scientific

No. of pages: 8