The traditional method of cell microscopy can be subjective, due to observer variability, a lack of standardization, and a limited feature set. To address this challenge, we developed an image classifier using a machine learning approach. Our system was able to classify cytoskeletal changes in A10 rat smooth muscle cells with an accuracy of 85% to 99%. These cytoskeletal changes correspond to cell-to-cell and cell-to-matrix interactions. Analysis of these changes may be used to better understand how these interactions correspond to certain physiologic processes.
Proceedings Title: 1st ACM International Health Informatics Symposium
Conference Dates: November 11-12, 2010
Conference Location: Arlington, VA
Pub Type: Conferences
digital image processing, machine learning, molecular biology