Deep Learning to Help Predict Patient Lifespan

2380
Deep Learning
Source: The Medical Futurist

Predicting the lifespan of a person is a difficult task. With so many factors at play, it’s nearly impossible for even the best doctors to make such a prediction. The best the doctors can do is make an educated guess using all the data at hand and their experience. However, there is something that can aid doctors to do that, and with much more accuracy. Artificial intelligence (AI) is right here for the doctors. At the University of Adelaide, researchers came up with a system that predicted the lifespan among 48 patients. This AI system discovered a pattern based on radiological test scans and predicted the patients’ mortality in five years. Researchers reported that the prediction was 69% accurate, a rate equivalent to that of a trained oncologist. Researchers plan to take this step further using deep learning.

Deep Learning Inspires an AI System

According to one of the researchers, genetic epidemiology professor Lyle Palmer, the system hopes to model itself after the human brain’s learning ability. The AI system can achieve this level of analysis through deep learning. Academia and industries are applying deep learning to myriad areas, thanks to the availability of large data sets. Palmer added that in medicine, deep learning is still in the early stages. But the research shows the computer’s ability to identify the appearance of diseases through complex images. This is a skill only well-trained doctors are able to do after years of training.

Going Beyond Predicting Lifespans

The research is also looking into using hospital images for more accurate prognosis and diagnosis. Once the AI system improves, it can predict diseases such as cancer and diabetes before the obvious symptoms arrive.

Currently, the system is still in the early developmental stages. Researchers are improving the AI system and doctors can’t use the information provided. The system is only able to predict a 5-year mortality within the small population. But the researchers are moving forward with their research, as the system now predicts death at an 80% accuracy.

This research hopes to also assess a patient’s lifespan beyond the doctor’s limitations, as they can’t look into the health of a person’s organs. Deep learning will certainly move this radiology-based research forward as more information is fed to the system.