Researchers at the Massachusetts Institute of Technology have demonstrated that robots can help with decision making on a hospital's labor and delivery floor.
Researchers from MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL) used a robot that could learn from human schedulers to anticipate bed flow and suggest which nurses to assign to C-sections and other procedures.
While robots long have been used to perform small tasks, such as delivering linens or medications, they’re now taking on higher-level responsibilities that can include surgery and doing rounds.
At Beth Israel Deaconess Medical Center, for instance, the labor unit head nurse has to coordinate 10 nurses, 20 patients, and 20 rooms at the same time, leading to an overwhelming number of possible combinations, according to an announcement.
The work was based on a simulation of a labor and delivery floor developed as part of a quality improvement program. Researchers then created a vision system in which the robot could “read” the whiteboard that served as the unit’s scheduling dashboard and use speech-to-text software to add in staff conversations around scheduling. The developers then created a decision-support system incorporating the two.
The staff don’t have to “train” the robot, according to the report--it learns by watching humans. The machine learning system looks at the actions humans make in various situations as well all the actions not taken as it creates a scheduling policy. It’s also able to respond dynamically to situations it has not encountered previously.
In the demonstration, nurses accepted the robot’s advice 90 percent of the time, but when it was programmed to give bad advice, they rejected that advice 90 percent of the time, according to the report.
However, nurses reacted positively to the robot--even more than they did for computer-based decision support. The nurses said the robot evened out the workload and helped new nurses better understand their duties.
Check out the video from MIT:
To learn more:
- read the research paper (.pdf)