MIT's algorithm searches physician's notes, extracts information from medical records

In the past decade, the physicians have adopted computers to store the medical data and related information electronically. These records are treasures of important information; for example, if searched properly it could reveal even more information about patterns between the symptoms, treatment used and the recorded outcome. This knowledge would help in future development of drugs and their administration in patients. However searching these records is a challenge in itself because of what the researchers call word-sense disambiguation. An example of this word would be 'discharge' which could mean a discharge of a fluid from body or discharge of a person from the hospital. The computer algorithm has to be intelligent enough to understand the context of these keywords to be able to perform an accurate search.



Researchers from MIT’s Computer Science and Artificial Intelligence Laboratory will present their research on this issue at the annual meet of American Medical Informatics Association’s (AMIA). The new algorithm developed by MIT's researchers can operate with 75% accuracy, making sense of the physician's clinical notes. MIT postdoc student Anna Rumshisky is leading the research and says that the team has taken a radically different approach towards the word-sense disambiguation problem. 'Topic Modelling' is the technique used by this team that identifies the topic of the document by inter-relating the keywords. An advantage of this approach is that it can be used on huge amount of texts without human intervention.

The team believes that their algorithm will be able to extract the information hidden deeply inside the medical notes created over the years. We won't be surprised that once the research is complete, Google executives will be ready with job offers to the entire team. Go Google!

Via: #-Link-Snipped-#