IBM's cognitive system will help doctors automate the analysis of medical imagery.

IBM in late November introduced a new solution for analyzing medical images based on IBM Watson. The solution was presented by Watson Health together with Merge Healthcare. The project is designed to improve the efficiency of doctors in the diagnosis, treatment and monitoring of patients. The system was developed based on the results of a decade of work by IBM specialists in the field of artificial intelligence and machine learning. As a result, the experts were able to “teach” Watson to analyze the images, drawing conclusions and learning in the course of work.

Images in medicine are an important source of patient health data. According to experts, the amount of this information in medicine is about 90% of the total volume of all data. In general, this is a good trend, but doctors in many hospitals are simply inundated with such images. Even the most experienced specialist can not always immediately understand the image, and in fact all this must be done quickly and accurately. In some hospitals, doctors have to look at thousands of shots daily.

Automated tools that would help doctors perform this work, while a very small amount. In most cases, doctors do everything manually, independently. There is not much time left for analysis, and this is the most important work, which largely determines the result of diagnosis and the course of treatment.
')
IBM Watson has already learned how to analyze such images, plus to everything, the cognitive system can, when studying an image, associate findings with data from a medical history, test results and genetic tests.



IBM specialists managed to teach the system the following:

• Analytical evaluation of images. This has already been discussed above - the system allows you to compare the results of the analysis of images with other data;
• Summarizing data helps radiologists, cardiologists, and other professionals use imaging analysis data in the diagnosis and treatment of patients;
• Detect bleeding - this feature helps emergency room doctors diagnose a stroke or bleeding in the brain. In this case, the “Brain Bleed” tool is used;
• Personalization - helps physicians use an individual approach to treating a particular patient, integrating the results of image analysis with other information relevant to the patient.

IBM specialists also developed a number of tools:

• Marktation — A service that processes medical images before doctors view them. This helps professionals work faster. The service will work first with mammographic images, then the scope of its use will be expanded;
• Watson Clinical Integration Module - a cloud application for physicians of radiologists, which helps to increase the efficiency of image analysis and reduce the number of errors;
• Lesion Segmentation and Tracking Module - this service helps radiologists effectively diagnose cancer patients.

In addition, doctors will be able to verify their diagnosis with a diagnosis of Watson. The cognitive system will process the received data in real time and make its own diagnosis.



Demonstration of IBM Watson capabilities at the RSNA 2016 conference ( Radiological Society of North America Annual Meeting )

System developers have already demonstrated its capabilities. In particular, IBM Watson has shown its ability to use data from thousands of sources to help diagnose the disease of several patients. The data provided by the service is not final - it is a summary report with conclusions that help a specialist to work - a person. In particular, IBM Watson helped identify anomalies - aortic dissection or pulmonary embolism.

In general, IBM Watson is able to help doctors by taking off part of their routine image analysis responsibilities, helping them to process and analyze such information. The ability to use additional data with the information obtained during the analysis of images helps to make an accurate diagnosis with a minimum probability of error.

Subject to the successful completion of clinical trials, a new service will soon be able to fully assist physicians in their difficult task.