How AI Is Changing Oncology
IBM Watson Health
The amount of research in healthcare is astounding and has surpassed that which anyone could consume with 8,000 research papers being made a day. One of the most powerful Narrow AI systems called IBM Watson can read 25 million research papers a week. During a recent interview on 60 minutes, Dr. Sharpless of UNC Cancer Center spoke of a recent project comparing recommendations made by a group of expert physicians vs. IBM Watson Health on 1000 cancer patients. First off Watson came to the same treatment recommendations as the tumor board in 99% of cases. The most interesting finding was that in 30% of cases it was able to give additional recommendations about clinical trials and genetics markers that patients could benefit from. This added information had been published within a week or two and not something many of the physicians were aware of.
IBM Watson Health has also been used to scan through tens of thousands of CT images to help discern normal structures from non-normal structures. This will helping radiologist diagnosis cancer, recurrences, blood clots, and many more. The other interesting use of IBM is uploading it to robots that can help patients and doctors with medical questions.
What is Narrow Intelligent AI: That means that is can make exceptional decisions within the parameters of the data it is given. We would not expect IBM Watson Health to give an answer about Fusion energy because the input data is only healthcare related publications.
What is a tumor board: These are meeting where all healthcare providers that specialize in cancer come together to discuss individual patients. To have a complete tumor board discussion one needs a radiologist, pathologist, radiation oncologist, medical oncologist, surgical oncologist. Genetic counselors, nurses, residents, and other staff may also be present.
Will this change how decisions are made in tumor boards?
Most likely yes, with the accuracy of the system and the additional recommendations for clinical trials and genetic markers, in the years to come making decisions without it will be considered suboptimal care. An analogy can be made with the National Comprehensive Cancer Network (NCCN) guidelines that oncologist have been references for years. It is an algorithmic powerpoint made by expert oncologist around the US. If that is now considered standard of care I have no doubt that IBM Watson Health will be as it evolves over the next several years. Currently there are 14 centers that have access to IBM Watson Health system for cancer care support. There are 16 centers that are partnering to help improve its imaging review skills. There will also be a specific program to match patients to specific clinical trials. Watson will be a huge help in this area as the new clinical trials are coming out all the time and they have several exclusions and inclusion criteria for patients to be able to enroll.
The weakness in the IBM Watson system is related to the data it draws on. All research is made by humans in a system that encourages positive results and discourages negative results. There are several statistical methods that can be used to make a paper sound more impactful and there are several questions that are not answered due to the risk of the data leading to a negative results. Evening with a significant risk of bias in research, it is still worthwhile to continue the process as it is the most objective way we know to increase knowledge.
I believe physicians should still review the results of IBM Watson as expert can read between the lines knowing when certain parameters of a given study were done for good reason or just to help get the paper passed the review board. Also Watson may know the exact clinical trial to increase the chance of survival for the patient, but the physician will need to speak with the patient about how they feel and if they want to go through the challenges that come with treatment. Adding a narrow AI system to cancer care will increase the objectivity of decisions made and constant review of the literature that patients faced with their own mortality deserve.
Looking forward to the future!