IBM Watson

From Clinfowiki
Revision as of 03:55, 24 January 2015 by Annathehybrid (Talk | contribs)

(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to: navigation, search

IBM Watson

Introduction

Artificial intelligence (AI) is defined as the capability of machines to think like human beings. The use of AI or expert systems in medicine has been evolving since 1970‘s. Some of the previous expert systems in health care were INTERNIST-I, ISABEL, CONSULT, PEPID DDX and Mycin. This post focuses on IBM Watson.

What is IBM Watson?

IBM Watson is an artificial intelligence system that uses machine learning and can process natural language. Its technology is based on machine learning. Machine Learning, is a branch of artificial intelligence, and is about the construction and study of systems that can learn from data. [Ref: 15]. In essence, IBM Watson is an application of natural language processing, knowledge interpretation, reasoning and artificial intelligence and uses IBM’s Deep QA technology [Ref: 16]. This technology enables Watson to churn through many terabytes of information and create rules which aids in informed decision making with varied levels of confidence. [Ref: 16].

Watson Technology

Previous expert systems like INTERNIST-I, ISABEL, CONSULT, PEPID DDX and Mycin failed to make an impact in healthcare due to following reasons [Ref, 9]

  1. Long time to enter data in codified format
  2. Scattered data across different departments
  3. Lack of vetted current information.

IBM Watson has namely three capabilities that help overcome the above shortcomings.

  1. Natural Language Recognition: Watson can recognize the nuances of the natural language and interpret data source and understand user queries.
  2. Hypothesis Generation and Evaluation: Watson, using advanced analytics capability can churn through huge volumes of data to evaluate the diagnosis and treatments.
  3. Evidence based learning: Watson is a flexible expert “learning machine”. The feedback from the user to Watson for each outcome is evaluated for future analysis.

Watson can ingest vast amount of structured and unstructured data at very high speed and sift through millions of patient’s records and other data resources from disparate sources in matter of seconds. IBM Watson collects data from literary works, encyclopedias, news articles, thesaurus, medical text books, electronic health records, websites, publishes articles, expert opinions, cohort studies, physician’s notes, biofeedback from medical monitoring devices, information from comment threads in online patient community and radiological imaging. All this forms the knowledge basis for data analysis. [Ref: 3]

How can Watson be used in Healthcare?

Vast amount of medical data is in unstructured format and is clinically relevant. Healthcare stakeholders would like to extract the knowledge behind the data to increase healthcare quality and efficiency.

Watson aims to be a medical advisor/trusted diagnostic tool to the doctor. This technology gives doctor a chance to look at patient history and feed in his/her observation data about the patient symptoms. Based on the query, Watson will go through medical publications, research papers, medical guidelines, EMR data, past doctor notes to come up with three best conclusions along with confidence level and supporting evidence. Watson could even suggest clinical trials that a patient can enroll. The doctor has a choice to ask for additional evidence or add limiting factor based on his/her clinical experience or patient preference. Watson is a clearly a decision support tool and is not a “decision making” tool- It will assist Doctors in making efficient decisions and will not replace the role of Doctors & Nurses.


Current Commercial Applications of Watson in Healthcare

1. Utilization Management Assistant (UMA): IBM tied up with Well point insurance in 2011 and released a decision support tool or Utilization Management Assistant for clinical staff to get pre-authorization for treatment based on patients insurance policy. This will considerably reduce the approval time and ensure timely care to patients. It is currently deployed with select number providers in Midwest USA and will expand to 1600 providers by the end of 2013.The commercial offerings are named WellPoint Interactive Care Guide and Interactive Care Reviewer.

2. Interactive Care Insights for Oncology: IBM tied with Memorial Sloan Kettering Clinicians and Well point Insurance to use Watson’s advanced analytics skills and vast experience of medical staff to provide ‘precision’ treatment plans based on patient preferences, updated & verified medical guidelines and vast array of published reference data. This commercial offering is made available on cloud for remote access and usage. The remote availability and accessibility of data in bidirectional fashion will lead to evidence based learning from user interaction and feeding of patient outcome data. [Ref: 3, 7]


Current Challenges and Future Developments

1. Data Quality: Watson is as good as the information fed into it. RCT (Randomized Clinical Trial) have been given higher weight in the IBM evidence ranking system. Based on the funding source for the RCT trials, biases and falsified information may creep into the publications that are referenced by Watson [Ref: 6]. Also, the author bias and their commercial interest are bound to cause “distortion” in Watson’s recommendation. Hence, it is very important to check the quality of data fed into the system from primary and secondary sources. The research papers are usually published many years after the original research. This may lead to obsolete data being fed into Watson. Also, illegible physician notes, opinion based blog entries are bound to reduce data quality.

2. High Cost of Implementation: IBM Watson is a costly capital investment along with high cost of training the staff and maintenance. Watson team is proposing the idea of lending out Watson’s data crunching capability as IaaS (Infrastructure as a Service) to reduce the cost.

3. Lack of Medical/Claims Fraud detection capability: Uploading vast amount of patient and medical onto an online system poses the threat of security breach. Watson’s is yet to address this challenge. [Ref: 4].


Conclusion

However, like any new technology development, we must thread the path with “ cautious optimism” and make informed decisions.

IBM Watson is poised to address some of the main challenges facing the health care industry. If implemented correctly and accepted by the medical fraternity, Watson can lower cost, reduce medical error, maintain consistency and fasten the time to treat by a faster decision making process and thus increasing healthcare quality and efficiency.


References:

1. Evidence Based Medicine[[1]]

2. Washington Post - Watson Patients Phyiscians[2]

3. IBM_Watson_will_eventually_fit_on_a_smartphone_diagnose_illness[3]

4. IBM Watson Supercomputer[4]

5. Brains Behind IBM Watson[5]

6. Computer Aided Medicine[6]

7. IBM Press Release[7]

8. Mc Kinsey Study on US Healthcare [8]

9. The Next Revolution in Healthcare[9]

10. D. Ferrucci et al., Watson: Beyond JeopardyD. Ferrucci et al., Watson: Beyond Jeopardy!, Artificial Intelligence (2012), [10]

11. Understanding-How-Big-Data-Flows-in-Healthcare-Infographic.jpg[11]

12. You Tube Video- Putting IBM Watson to work[12] 13. Machine_learning[13] 14. The_Big_Data_Revolution_in_Healthcare__June_2012_.pdf[14] 15. Health_Information_Technology_for_Economic_and_Clinical_Health_Act [15] 16. Watson Computer [16]


Submitted by (Binitha Surendran)