THE FUTURE: Improving Diagnosis of Lymphoma and Leukemia Using Image Understanding Systems
Author:Hernani Cualing, M.D., Assistant Professor of Pathology and Laboratory Medicine (Hematopathology).
To date four major technologic developments have dramatically altered the practice of pathology: Light microscopy, electron microscopy, immunophenotyping and molecular biology. A fifth technique is now emerging which has the potential to alter practice once again. It is digital image analysis, which has been evolving since the 1960's to the point in the 90's where the interface with artificial intelligence is becoming a reality. Much of pathologic diagnosis is based upon manual visual techniques and automation of such processes using computer assisted imaging techniques has the potential to improve the speed and decrease the cost of diagnosis. We are currently evaluating the application of this technique to the diagnosis of lymphoma and leukemia which together are the third most rapidly rising cancer in the United States. Pathologists correctly diagnose lymphomas on initial cytology in 51% of cases and this cytologic classification is identical to the histologic classification in 85% of cases (1). Ancillary tests such as immunophenotyping and molecular genetics enhance the ability of the pathologist to characterize malignancies beyond morphologic diagnosis but these tests are complicated, costly, and time consuming. The addition of image analysis to this process has the potential to eliminate the need for most ancillary testing making diagnosis both more rapid and less costly. In cooperation with The Artificial Neural Systems Laboratory in the UC College of Engineering, we have developed an algorithm for lymphoma/leukemia diagnosis which combines digital image analysis with neural network analysis to identify and classify cells. Slides of patient tissues prepared from cytologic smears, imprints and fine needle aspiration can be used. The project is at the prototype stage but based upon preliminary data, it appears that application to clinical use will be possible in the near future.