by Susan Eustis
IBM Watson has been positioned to leverage a known knowledge base to assist in cancer diagnosis. Facilities that are leading in their field provide direction for the automated process computer. Computer-assisted surveillance will play an increasingly important role the in helping clinicians diagnose cancer conditions.
Melanoma is treated at Memorial Sloan-Kettering Cancer Center in New York City. Clinicians there are interested in how Watson can be used to provide a more consistent approach to care delivery across the entire medical spectrum. Watson technology has the potential to help clinicians identify suspicious lesions in a more consistent manner, and act as an aid to achieving diagnosis of an individual concerning lesions. Melanoma and cutaneous malignancies are matters of interest.
Watson, an IBM computer represents a change in computer evolution. It can understand natural language. Based on this understanding, it can query a huge database to come up with a clinical diagnosis.
IBM is developing the ability of Watson to interpret radiological images. The system is learning how to diagnose melanoma images. It needs to be taught in the same manner that medical students need to be taught. There are other systems that can assist in identifying and diagnosing lesions and these support the clinician and the decision support computer systems.
The value of Watson is that once it learns how to achieve a diagnosis, it will be able to be updated once and used millions of times. It is impossible for clinicians to keep up with every possible piece of information in the medical literature. Diagnostic tools like Watson can scan known conditions and can point out differential diagnoses that a clinician might not have considered. Using computers to assist in diagnosing melanoma and other cancers depends on having the computer trained at a leading medical center. .
In terms of identifying images and following changing lesions, a computer would need pictures from every patient visit. The aim is to create a system in which a patient walks through the door and gets a picture taken as part of the visit to the clinician.
Over time, one problem with using multiple images is registering the images in a manner that permits comparisons. The difficulty occurs because patients’ bodies change. Curved surfaces and 2D images, make registration and comparison difficult. 3D images solve this problem, making registration easier. In this manner computers are being trained in much the same manner that medical students are being trained.