Why Medical Doctors Should Learn AI?
The world of medicine is changing. We are seeing a rapid growth in the body of research of AI technologies, and AI technologies are being implemented in the medical practice more than ever. For example, one company, AetherAI, is implementing image recognitions to help pathologists to diagnose different disease. Some of its product are FDA-approved, meaning it’s about time to see its product being used in clinics. The WHO has even recommended the use of AI-based software to screen TB (but please note it’s a “conditional recommendation, low-certainty of evidence”. Read the full guidelines here)
But these changes seems to be relevant for engineers, right?. Why should medical doctors care about it? In this article, I would tell you why medical doctors
Why M.Ds should learn about AI
Even after autopilots in airplanes are the standard, pilots are not disappearing. Pilots are still needed to make sure everything runs smoothly, and handle lots of task which can not be automated. The same is the case with medical practice. Even if AI will be implemented more in medical practice, medical doctors will be the main decision makers to determine the patient’s treatment.
The second reason is that medical doctors may need to know the tools that helps them. Just as medical doctors should know about the basic knowledge of radioactivity to read radiographs, medical doctors should know about the basics of AI to be able to make any decision. AI technologies (from automated radiograph reading, prediction for disease severity to automated diagnosis) still has a lot of bias. By understand the basics of how AI works, medical doctors can gain benefits in using these new technologies.
Lastly, medical doctors will also be the key player to determine whether they want implement the AI technologies to their own practice. We have seen a lot of digital health and AI models being thrown out because the clinicians do not want to adopt them (because some digital technologies are putting more works, or requires too much resource, etc). Just as medical doctors may not directly involved in the production of drugs in the factory, they are having a say whether they are going to prescribe the drug. In this case, it is appropriate for medical doctors to understand the basic metrics of the evaluation of AI model performance.
AI is coming, but stay skeptical
Although implementations of AI can be helpful to clinicians and to patients, we have to admit that there are a lot of downsides as well. AI requires data, and this can pose risk for data security and privacy. AI models can be deceiving as well, and may even recommend false treatments. Most troubles are coming because people are expecting too much about the technologies, or because people resisting since they are reluctant to adopt new useful technologies. The most appropriate way is by knowing the AI we are trying to use, without any judgement to overuse or underuse it.
In this case, medical doctors may need to understand the AI itself, and not biased with their own bias to new technologies. Even though most machine learning engineers are trying to handle the “black box problem”, and making the machine learning technologies to be interpretable to every stakeholders, some knowledge about the basics of AI and machine learning will help clinicians a lot.
Also, before implementing any new technologies, we need to rely on the evidence. We need to be skeptical and assess the technologies in clinical, economical, political, technical and social aspects.
Knowing the background and technicalities of the AI system will be the only way to ensure we as doctors can deliver the best healthcare possible. And of course, to do no harm.
Hello! Thank you for reading this article. Please help me to create better writing for you by commenting on your inputs. If you find yourself liking my writings, feel free to leave an impression, clap, and follow my profile. Help your friends to find me by sharing this article. Hope you have a good day!