Recently, researchers from South Korea have developed an AI doctor that can automatically obtain needed patient records and data to provide an accurate diagnosis. Why would AI doctors benefit patients, what did the researchers develop, and what challenges do AI doctors face?

How AI doctors would change a medical diagnosis

Many advances in silicon technology, the reduction in cost for computing, and the immense amount of data being produced globally have meant that AI as a technology has only become practical in the past decade. Unlike software algorithms of the past, AI allows for an application to be presented with new situations that it was not specifically programmed for and then find a valid solution for that situation.

For example, a self-driving car with cameras and sensors could be presented with a new stretch of highway that it has never seen before. With AI, a self-driving vehicle could successfully navigate the road (thanks to the past experience from other routes that it has successfully driven).

One area that AI could prove to be invaluable is the medical diagnosis industry. Diagnosing medical conditions accurately requires a doctor to have access to a patient’s entire medical history, a deep understanding of medical science, and the ability to form links between a patient’s symptoms and potential diagnosis.

While human doctors may remember the vast majority of a patient’s medical history from one reading, they will never be able to recall all medical knowledge, including research papers, journals, and data from other patients. As such, doctors will either make false diagnoses or order multiple tests to try and narrow down the list of potential conditions.

However, imagine an AI doctor that could do all of this with perfect memory recollection while also analysing millions of other patient records and research papers. Simultaneously, this AI doctor would be serving thousands of patients using cloud-based technology while providing a service whose costs would be a fraction of that offered by modern medicine.

Furthermore, the doctor interface could be accessible via a standard web browser, and no wait times to see a doctor could see conditions diagnosed in a much shorter. Removing the need for a physical GP would further allow human doctors to heavily specialise in one particular area, which would increase the number of specialists available to patients.

Researchers develop AI doctor services with promising results

Recently, researchers from South Korea developed an AI doctor that can diagnose cardiovascular conditions in patients with an accuracy close to 90%.

For the new system (currently in development) to operate, it requires access to patient records. Under normal circumstances, patient data has to be carefully monitored and shared as it contains sensitive information on patients. Patients are generally required to sign documents to allow these documents to be shared amongst various hospitals and departments.

In order for the AI to function, it requires access to medical records, and these medical records have to be in the form of an Electronic Medical Record (EMR). As such data is extremely sensitive, researchers from the Electronics and Telecommunications Research Institute (ETRI) have developed a new AI system that can collect such data from hospitals and coordinate with other AI systems to help create diagnoses.

Furthermore, the new system helps prevent sensitive data from being exposed while only sending data that is absolutely essential for the diagnosis. The current system has shown effectiveness with diagnosing cardiovascular conditions by looking at over 740,000 patients and correctly predicting over 90% of the cases.

What challenges do AI doctors face?

AI doctors will undoubtedly prove to be a handy tool for future medical diagnosis, but their use must be carefully regulated with checks and balances put in place. The vast access to private medical records could make doctor AI systems a target for cybercriminals who want to mine personal data for identity theft.

Authoritarian governments may also look towards electronic medical systems for patient data as a way to monitor citizens. While this may seem unlikely, the truth is that the UK government is already planning to introduce legislation that would allow the police to look at patient data held by the NHS to look at crime statistics.

Once this data access is given, it is more than likely that the authorities will use it to track individuals who request medical assistance, which could be a significant breach of doctor-patient confidentiality. This is already a reality in places such as Dubai, where patients in hospitals who have traces of illicit drugs in their system are reported to the police.

Overall, AI doctors could provide the entire world with a diagnosing system that is far cheaper than human doctors, give better results, and be instantly accessible. However, we must be careful how such systems are implemented, and above all else, must ensure that authorities and governments are denied access to such data.