Artificial intelligence is becoming more and more relevant in healthcare matters. As I mentioned here, Ebola outbreak was controlled only after artificial intelligence was introduced in the campaign against the epidemic. So, what is artificial intelligence applied to healthcare? The idea is to use computers to facilitate prevention and interventions against diseases. These tools are built through machine learning, which is the development of algorithms and statistical models that computers use to resolve a healthcare-related task.

How does a computer learn to do diagnosis?

I will explain it with an example. Thousands of health-related images (e.g., glass slides of biopsies) are submitted to a computer with related information and data. The computer “learns” to associate the data to the image and after this training it will be able to derive data by itself from random slides. So the machine will be able to diagnose a subtype of cancer from a particular staining of the biopsy on the slide, for example. Is the machine always right? No, unfortunately, but it doesn’t make more mistakes than a pathologist. The benefit of using machines is that in cases of emergencies (like for Ebola outbreak), diagnosis time is reduced, and work is more processive. Clearly, in medicine, we are very far from having machines able to replace human.

The role of artificial intelligence in healthcare is still relatively limited, but the potentialities are paramount, as it was demonstrated by Liang and colleagues in an article published on Nature Medicine in February 2019. The topic is complicated, so I will try to summarize it in a simplified way. They exploited artificial intelligence to predict pediatric diseases based on electronic health record of children. Deep machine learning was used to tech the computer how to extrapolate these data and how to select the relevant ones. 101.6 million data points from 1,362,559 pediatric visits were analyzed to train and validate the machine. The result indicated that artificial intelligence was then able to diagnose pediatric diseases with the same efficiency of expert pediatricians.

This success demonstrates that artificial intelligence support can help in tackling large amounts of data, augmenting diagnostic evaluations and providing clinical decision support in cases of diagnostic uncertainty or complexity. Artificial intelligence is already successfully used in radiology, dermatology and ophthalmology. Let’s expect in the close future to see a more prominent role of computers in supporting doctors.

References

Bempong NE, Ruiz De Castañeda R, Schütte S, et al. Precision Global Health – The case of Ebola: a scoping review. J Glob Health. 2019;9(1):010404.

Liang H, Tsui BY et al., Evaluation and accurate diagnoses of pediatric diseases using artificial intelligence. Nature Medicine. 2019 https://doi.org/10.1038/s41591-018-0335-9