What even the most acclaimed professors know cannot match cognitive computers. As the amount of information they accumulate grows exponentially, the assistance of computing solutions in medical decisions is imminent. While a physician can keep a few dozen study results and papers in mind, IBM’s supercomputer named Watson can process million pages in seconds. This remarkable speed has led to trying Watson in oncology centers to see how helpful it is in making treatment decisions in cancer care.

Watson is based on deep Q&A technology and gives a set of possible answers as the most relevant and likely outcomes to medical questions. But physicians make the final call. I have to note here that Watson is not there to replace the physicians, but to support them when making decisions. It also interacts with physicians and can suggest which additional tests are needed to generate a higher degree of confidence.


The MD Anderson Center’s Oncology Expert Advisor

It is built to aid physicians in making evidence-informed decisions based on up-to-date knowledge. The system was designed to have three main capabilities:

  • Dynamic patient summary: Interpret structured and unstructured clinical data to create dynamic patient case summaries.
  • Evidence-based treatment options: Make treatment and management suggestions based on the patient profile weighed against consensus guidelines, relevant literature, and MD Anderson expertise.
  • Care pathway advisory: Provide care pathway advice that supports management of patients by alerting clinicians of adverse events or suggesting proactive care support.

When testing the accuracy of the system to recommend standard of care treatment related to 200 leukemia cases, the system had a false-positive rate of 2.9% and a false-negative rate of 0.4%. The overall accuracy of the standard of care recommendations was 82.6%.

The Memorial Sloan Kettering Oncology Advisor

Memorial Sloan Kettering’s expertise and experience with thousands of patients are the basis for teaching Watson how to translate data into actionable clinical practice based on a patient’s unique cancer. While initially focused only on breast and lung cancers, the work has expanded to more than a dozen other common solid and blood cancers such as colon, prostate, bladder, ovarian, cervical, pancreas, kidney, liver, and uterine, as well as melanomas and lymphomas. Watson digested the guidelines about Lung and Breast Cancer issued by the National Comprehensive Cancer Network (roughly 500,000 unique combinations of breast cancer patient attributes; and roughly 50,000 unique combinations of lung cancer patient attributes). Over 600,000 pieces of evidence were digested from 42 different publications/publishers.

How to prepare

There is no doubt it will have a bigger and bigger impact on how we practice medicine worldwide. But all stakeholders in the system must prepare for that:

  1. Medical professionals should acquire basic knowledge about how AI works in a medical setting in order to understand how such solutions might help them in their everyday job.
  2. Decision makers at healthcare institutions should do everything to be able to measure the success and the effectiveness of the system. This is the only way to assess the quality of AI’s help in medical decision making.
  3. Companies such as IBM should communicate even more towards the general public about the potential advantages and risks of using AI in medicine.
  4. Non-English speaking countries should invest in natural language processing (NLP). If the patient information is not in English, Watson needs to understand the content and context of the structured and unstructured information in that language. To do this, it uses the Unified Medical Language System (UMLS) and a semantic type recognition. The Watson Content Analytics (WCA) tool that processes NLP and is based on Unstructured Information Management Architecture (UIMA) is used for building annotations. WCA then uses a Medical Concept Extraction Tool and a Health Language Medical Terminology Management system that uses standard medical terminologies databases such as SNOMED, ICD-9, ICD-10, RxNorm, etc. And this is where most e.g. European countries miss the point. They don’t have all these systems in all the languages.

The other option is obviously to train physicians and nurses to document everything in English. But we can agree that this will never happen.

It is time to prepare in order to let technology help us do a better job in medicine.

This video provides a great summary about all these: