AI medical diagnostics: Open discussion by data scientist, medical expert, and ignorant traveler

AI medical diagnostics: Open discussion by data scientist, medical expert, and ignorant traveler

How close is the mankind to the capability of making automated and accurate medical diagnoses? The troubles I faced during my last vacation revealed that such reality is not as near as it may seem to be. But to figure everything out, I asked two healthcare professionals – a medical practitioner and a medical expert – to have a discussion. And I also invited a data scientist to comment on the matter.

So, how did the story begin? Everything started with a trip around Spain suddenly broken by my feeling ill. I was surprised to find out new shades of insurance medicine when I had to see a medical specialist. The doctor carefully listened about my symptoms and diligently entered the data into a computer program. The system reacted by providing a most likely diagnosis and a recommended medication treatment. ‘That’s the local insurance company line,’ the doctor explained.

In fact, the work of a local medical practitioner was restricted by the capabilities of an expert system. On the one hand, it eliminates medical errors that lead to the excess of insurance coverage. On the other hand, it puts significant restrictions on making a diagnosis by a medical specialist. My case was a confirmation of such frustrating limitations: the medications prescribed by the system didn’t help. My second visit to the doctor would be purposeless as well, if not for my request of close medical examination for sunstroke. By the way, my new medication treatment was far more effective and allowed me to continue my vacation time.

Medical Expert. From the perspective of that doctor, who examined you, the symptoms could be interpreted as a viral infection. Sunstroke is a symptomatic condition that the system didn’t take into account since it requires emergency help. For sure, there were many overlapping factors that distorted the real situation and prevented the doctor from making a right diagnosis. Despite the use of an expert system, the doctor always has the final word. I believe, if you had timely applied for medical help, the doctor would not have made that mistake. From the medical statement, we can see that you visited the doctor late at night, and that is why he didn’t link your symptoms to sunstroke. The expert system is also to blame: it doesn’t take into account a number environmental factors as temperature, humidity or the season when a disease progresses, but such advanced systems already exist. As for the doctor, he, actually, wasn’t diligent enough and cared more about acting within insurance limitations and receiving the insurance payment for your visit than about your health condition. It is an acute problem of insurance healthcare provision.

Indeed, insurance costs were the problem. My insurance company refused to pay for the second visit, explaining that sunstroke case is not covered by insurance and leaving me no chance to dispute the matter. On the other hand, I received low-quality medical help which jeopardized my health. As a patient, I’m more interested in whose fault was it and how to avoid such mistakes in future. Maybe the expert system is the problem, and it is better to apply tried and true methods to solve diagnosing problems?

Data science specialist. I am sorry for your case. I agree that a number of factors led to that mistake. However, it is not the reason to reject technical progress and any expert medical systems. As you mentioned, they help to resolve a number of issues, in particular, the compliance with insurance requirements, implementation of up to the date treatment methods and medications. It may take time to retrain medical personnel, but updating medical expert system’s knowledge base will have a global nature. For instance, recently, the vaccine against rotaviruses has been developed, and those healthcare facilities, where they employ expert systems, have been instantly updated on the new medication. They started to prescribe it to travelers and vacationers since the running expert systems took into account indications to use this medication. The other side of the coin is that an expert system may provide a probable diagnosis or – as in your case – may omit crucial factors. On this point, we come across the definition of an expert system – the system that assists in decision making. It is not an independent diagnostician, it simply assists medical professionals to gather extra information, compare it with the insurance requirements and make the right decision. Therefore, it is the doctor who makes the final decision, and he or she is fully responsible for the consequences.

So, can we say that today, automated medical diagnosing is out of the question?

Medical expert. Not quite. Even today we have plentiful examples of automated medical diagnosing by expert systems. Take a look at the diagnosing equipment: it requires blood chemistry, blood composition and properties to provide a most likely diagnosis and assist a medical professional in decision making. Take the electrocardiograph machines that detect many deceases and cardiac impairments. On the other hand, the final diagnosis should come as the result of complex diagnosing and patient monitoring by an experienced medical professional.

– And what about AI and self-learning neural networks? Can we train a neural network to consider various medical histories and make an automated diagnosis? 

Medical expert. Unfortunately, not. It is expected to function so in the far future. But if we speak about some common medical cases, such approach may work. Self-learning neural networks will be capable of spotting seasonal diseases and well-diagnosed health problems. But when it comes to probable diagnosing, the accuracy of automated processes will decrease. I believe you wouldn’t trust a diagnosis that neither doctor nor data science expert is certain about. The main problem for the developers of self-learning expert systems lies in the quantity and quality of learning data. In medical practice, we talk about medical cases, results of analyses, strategies for patient treatment and monitoring. For each of these points, there are certain inaccuracies and errors. For example, a doctor formulates a medical history in his own way and doesn’t comprehensively fill in the form or overdo it. Also, different equipment may be employed for medical testing which results in different errors or confusion of the analyses. Given that a patient strictly follows doctor’s prescriptions, he or she may actually use analogs for medications or generics… How can we monitor each stage and make sure that one or another medical case is true? As I mentioned before, each stage may contain false data, and when overlapping, the factors result in more and more errors decreasing the accuracy of data.


As for now, e-medicine is not ready to “treat” patients in the automatic mode. Medical professional’s experience and expertise is still the cornerstone of effective patient treatment. To date, many diagnosing areas lie within the coverage of automated or semi-automated expert systems. Data science specialists still have many issues to solve before expert systems will be able to make automated diagnosing. But this future hasn’t arrived yet.


Katrine Spirina

Marketing Analyst


Phone: +375 (29) 298-36-28

Please submit a short project request, and our account managers will get back to you within 24 hours to discuss our future cooperation.
Start Your Project with OCSICO