Medical imaging has been rapidly advancing over the last few decades. Non-invasive acquisition of clear and accurate images of the internal organs and tissues has aided in the diagnosis of many diseases, in some case even without the need of biopsy. With the advancement of technology histopathology has seen strides in the digital realm. Histopathology is the study of changes in tissues caused by disease, and now with the use of a CBIR systems (content-based image retrieval) these pathological examinations can be done more efficiently and accurately than ever.
The University of Waterloo is home to innovators looking to accelerate the applications of technology in healthcare and well-being. Centre for Bioengineering and Biotechnology Member Dr. Hamid Tizhoosh, the director of Kimia Lab, at the University of Waterloo and his main research areas include computer vision and machine intelligence. He also has extensive experience in medical imaging.
The problem with pathology today is that to complete a diagnosis, the opinion of a second pathologist is often conscripted. This process is long, expensive and overall inefficient. Under Dr. Tizhoosh’s supervision are Habib Herbi and Wafa Chenni, who are here as part of the student exchange program between Sorbonne Universities and the University of Waterloo. They have been helping look at ways to create more accurate representations of digital images of tissue samples and to devise a way for quick identification of this samples.
They decided to place these images into a CBIR system for fast comparison to existing tissue samples in the database. The CBIR system works by using an image database to extract and match features of a query image. This will be done for the purpose of achieving more efficient classification and reliable pathological diagnoses. This method of using machine learning to teach a system to recognize images has the potential to eliminate the need for a second pathologist’s opinion during classification.