IBM Watson Health completed three years in business. Is the supercomputer technology living up to the audacious expectations, IBM created for it in the beginning? Sandeep Makhijani, Watson Health Leader, Asia Pacific, IBM Watson Health reveals their progress and plans with Raelene Kambli
What is IBM’s vision towards solving the most pressing health challenges through data and AI insights?
IBM launched Watson Health three years ago with a clear goal of find the best ways to bring AI and other technologies to help medical and health professionals tackle the world’s biggest healthcare challenges. We are continuing to improve the ways that Watson Health technologies can be integrated into clinicians’ workflows and existing processes, as well as continuing our focus on scientific research and localising products. We’re continuing to focus on our efforts in this area to build trust and ensure our technology is delivering clinical value.
Tell us about Watson’s achievements in India so far?
We have collaborations with many hospitals in the Asia Pacific region, (which includes India) as well as across the world for implementing IBM Watson Healthcare solutions. In India, we have two leading healthcare entities – Apollo Hospitals and Manipal Hospitals – who have adopted Watson for Oncology. At Manipal Hospitals, for example, the multidisciplinary tumour board uses Watson for Oncology to review challenging cases, and they have found that the technology is 92 per cent concordant with their multidisciplinary tumour board, according to a study in the Annals of Oncology.
How does the technology actually work for Indian physicians and patients?
Emerging technologies like AI are set to transform the healthcare sector in India. From hospital care to clinical research, AI applications are changing how the health sector works to reduce cost and improve patient care. At Watson Health, we have impacted over 15,000 clients and partners, 80,000 professionals and 185,000 patients and consumers with our AI solutions.
For example, Watson for oncology is trained by oncologists at Memorial Sloan Kettering Cancer Center, and is a cognitive computing system that uses natural language processing to ingest patient data in structured and unstructured formats. The system provides physicians with treatment options that are derived from established guidelines, the medical literature, and training from patient cases. Watson for Oncology is continuously learning over time, and doctors have access to peer reviewed studies, clinical guidelines, and expert perspectives.
Similarly, IBM Watson for Genomics analyses massive bodies of genomic, clinical and pharmacological knowledge to help uncover potential therapeutic options to target genetic alterations in a patient’s tumour. Using this genomic analysis, Watson produces a report for physicians, which identifies genetic alterations that are actionable based on literature as well as drugs and clinical trials that target those alterations.
Is the technology exclusively based on training by human overseers, who feed Watson information about how patients with specific characteristics need to be treated? If yes, what happens to those rare cases?
IBM Watson for Oncology is trained by Memorial Sloan Kettering (MSK) to compliment the work of oncologists, supporting them in clinical decision-making by enabling them to access evidence-based, personalised treatment options from more than 300 medical journals, more than 200 textbooks, and nearly 15 million pages of text providing insight and comprehensive details on different treatment options, including key information on drug treatment selections. The system has been trained to support treatment of 13 cancer types which represent 80 per cent of the global cancer prevalence. The initial treatment options based on the MSK training are generated for each patient with the associated reference material such as the textbook and medical journals etc. The physician then determines and discusses the appropriate treatment with the patient. Usually, when a cancer patient visits an oncologist, the doctor based on his experience, knowledge of the cancer type, various clinical diagnosis of the patient determines the treatment plan which typically includes drugs, radiation, chemotherapy and others.
What happens to the system if the historical data of patients is not well organised by hospitals? How will Watson react to such crisis?
Watson for Oncology is capable of identifying the key attributes from a patient’s record. If information that is important for treatment decisions is missing from the record, the treating oncologist will be prompted to enter that information into the system manually before Watson for Oncology can provide treatment options. Ultimately the treatment decision is always up to the doctor and patient.
Has IBM published any scientific papers demonstrating how the technology affects physicians and patients?
We have published more than 50 studies on our Watson AI technologies, and more studies will be coming out this summer. In June 2018, the latest study of Watson for Genomics was presented at the annual meeting of ASCO (American Society of Clinical Oncology), where the Guangdong Lung Cancer Institute in China found that Watson for Genomics matched the bioinformatics molecular tumour board’s manual analysis of mutations in 43 per cent of lung cancer cases. But in the other 57 per cent of cases, Watson for Genomics found 1.54 additional mutations, on average, that the bioinformatics molecular tumour board had missed (n=115). With that additional level of accuracy in the genetic alteration of a tumour or lung cancer, the targeted therapy can be further fine-tuned by the clinician for the patient. We have done several such clinical studies which show that there is a very good concordance for both Watson for Oncology and Watson for Genomics.
In another study, disclosure of Watson for Oncology recommendations resulted in prescriber treatment changes in five per cent of cases. The adherence rate in the 106 cases where decision changes were made improved from 89 per cent to 97 per cent (ASCO 2018 http://abstracts.asco.org/214/AbstView_214_229459.html)
Major hospital chains such as Apollo and Manipal have tied-up with IBM Watson so far? What has been their experience, challenges and benefits so far?
Apollo Hospitals implemented Watson for Oncology and Genomics to help physicians provide patients with personalised, evidence-based cancer care. This agreement was the first-of-its kind Watson for Oncology and Watson for Genomics deployment in India. The solutions also helps oncologists at Apollo surface relevant data to bridge disparate sources of information and identify treatments that are personalised to each unique patient.
Manipal Hospitals have been using Watson for Oncology to help identify evidence-based cancer care options for their patients. The benefits of man + machine are clear to the doctors. According to studies from Manipal, in 93 per cent of breast cancer cases, Watson matched the recommendations of the hospital’s tumour board —a group of 20 physicians who typically study each case for a week and spend an hour discussing it. It’s been fantastic for us to be collaborating with them and calling them our clients in India.
Various reports and investigations in the US, claim that that the supercomputer (Watson) isn’t living up to the lofty expectations IBM created for it. What is your opinion on the same?
Across the various areas of Watson Health our efforts are having an impact.
In just three years:
- Watson Health has more than 15,000 clients and partners.
- Watson Health cognitive offerings have impacted care or social services for more than 295,000 people.
- More than 50 peer-reviewed publications, posters, and abstracts support Watson Health cognitive offerings and 500+ pieces of scientific evidence demonstrate how our AI data and analytic tools are being used by clients and partners in healthcare and life sciences.
- IBM has more than 2,500 active granted and pending US patents in healthcare and life sciences. 400 of those are specific to Watson Health.
raelene.kambli@expressindia.com