Dr. AI: Challenges and Opportunities on the Road to AI-enabled Healthcare

AI-enabled technologies are simultaneously extending, complementing and jeopardizing global healthcare efforts. The road ahead for "Dr. AI" looks precarious yet promising.

446
A doctor with arms crossed holding a stethoscope
Illustration: © IoT For All

Healthcare systems have evolved rapidly during the last decade and are now providing a variety of benefits to patients throughout the globe. This is because the healthcare industry has embraced artificial intelligence and is currently utilizing its many applications to provide a better and safer experience to patients suffering from different ailments.

However, while discussing the benefits AI brings to the healthcare industry, one shouldn’t forget that all that glitters in’t gold. According to Bob Kocher, MD, an adjunct professor at the Stanford University School of Medicine, “if we are not careful, AI could…unintentionally exacerbate many of the worst aspects of our current healthcare system.” This doesn’t mean the advantages of AI should be ignored. This article analyzes and compares the various pros and cons of AI to the healthcare industry to arrive at a more comprehensive decision as to whether and to what extent AI should be made a part of this industry.

How AI Is Improving Healthcare and Who Is Doing What in the Field

From helping doctors with diagnostics to categorizing patients data to even assisting in surgery, there can be no doubt that AI has proved invaluable to the healthcare industry. What’s more, AI can also learn to identify good target proteins and suitable candidates for clinical trials to speed up the process of drug discovery and development. It’s no wonder then that big AI companies working in the healthcare space are constantly striving to improve their services to provide further facilities to the healthcare system. Here are some reasons why AI should be made a part of every healthcare system across the world.

Making Healthcare Accessible

In countries without robust healthcare infrastructure, accessing care is far from easy, especially for those living in remote areas where no healthcare facility is available. AI can be used here to create a kind of digital healthcare infrastructure or fabric in these regions, to provide not full-service care but at least some way of helping patients understand their symptoms and find appropriate treatment. Applications like Ada are providing health guidance to underdeveloped populations. Such solutions are free and available in the native languages of regions where they’re launched, increasing accessibility.

Predicting Diseases

Because AI has the potential to store all of people’s data in one place, it can access this information to compare patients’ previous ailments with the current symptoms to come up with a more accurate diagnosis. Since such apps have millions of the previous diagnosis stored, patients utilizing them don’t feel the need to go for a second opinion. Plus, by combining and analyzing data from various sources, AI can predict health problems that a person can get in the future.

One such application is Verily by Google, which started as Baseline Study and is working on forecasting non-communicable diseases like cancer and heart attacks as well as hereditary genetic illnesses. The aim is to enable doctors to predict any condition a person is likely to suffer in the future so that they can come up with treatment plans to avoid it or treat it in time.

Another way AI can help with diagnosis is by identifying biomarkers. Biomarkers are special molecules present in bodily fluids that can locate the presence of a particular disease in a person’s body. Because AI can automate the majority of the manual work in recognizing these biomarkers, it can save a lot of time and energy when it comes to diagnosing a disease. Since AI algorithms can skillfully classify molecules to identify a specific condition, they’re more cost-efficient as well; people will no longer have to go through expensive lab tests and other procedures like whole genome sequencing if doctors utilize the aid of AI in diagnosing diseases.

Assisting in Surgery

AI is now being used in operating rooms because applications based on artificial intelligence have proved to be a great help to doctors during surgical procedures. Robot-assisted surgery is now a thing of the future in the healthcare industry and has paved the way for successful surgical treatment of rare conditions. Complex operations can now be performed with precision with minimal side effects, less pain and blood loss and a quicker recovery through robot-assisted surgery, as reported by Mayoclinic.

Along with this, AI has now provided surgeons with real-time information about patients currently under treatment. This includes visual scans revealing the division of the brain into its various portions as well as MRI scans and other imagery required to assist in the operation. By laying these images on top of the patients’ body, AI has given X-ray vision to the doctors. This comes as a great relief for the patients who feel more secure when handing themselves over to their physicians under anesthesia.

Challenges Faced by AI in the Healthcare Industry

A common mistake that healthcare professionals make when incorporating AI into their system is focusing only on the advantages while ignoring the risks that the same system can bring. And there can be several potential drawbacks of relying only on machines instead of humans to help with a malady. Therefore, it’s necessary to have full knowledge of any AI system that a healthcare platform intends on using. AI itself has a long way to go before it can be unhesitatingly incorporated by any healthcare system. Below are some of the challenges AI faces in the healthcare industry and some of the things that can go wrong when, instead of finding ways to overcome these challenges, healthcare professionals depend on AI alone.

Limited Technology Makes for a Defective Diagnosis

As mentioned above, to make the correct diagnosis for any specific disease, AI is dependent on a variety of data gathered from millions of people who have suffered through a similar condition. Along with this, there should be sufficient data of such patients belonging to a particular group in AI databases to make the correct comparison. Hence, if there’s not enough data of patients from a particular background, AI will make an inaccurate diagnosis and doctors might make the mistake of going along with it if they aren’t experienced enough to recognize it as false.

To cite one example, in a study published by MIT News, skin-type biases were found in three commercially released artificial intelligence face-analysis systems. In dark-skinned women especially, the error rates increased up to 34% in two of the programs. This can pose serious threats when diagnosing and treating for serious skin conditions like melanoma on dark versus light skin.

Perpetuating Prejudices

Because machines based on AI lack compassion and understanding of human nature and conditions in which they survive, they often make economic and social biases when diagnosing patients. While these biases are mostly always unintentional, as emphasized by Dr. Rebecca Pearson, chief technology officer at ThoughtWorks, they play a great hand in sustaining such biases that are already causing disparities in the health sector.

For example, if a patient comes from a low-income neighborhood, an AI algorithm can take it as an indicator of poor social support system and recommend a nursing facility for treatment. Since this is more expensive than a home-based treatment, the patient may get discouraged to continue it. Only a doctor can look into the patients’ family and household environment to determine what kind of a treatment plan he can afford and is best suited to his individual needs. This is why Quartz recommends AI application developers to have better textbooks, i.e., to study the world and everyone in it, without class, racial and gender biases if they want AI to replace doctors.

Encroachment of Data Privacy

When dealing with AI applications, one needs to remember that they’re dealing with machines. And machines can malfunction! AI applications have provided the facility of storing all of the patients’ information. This includes records of all previous ailments as well as personal information and medical reports of blood tests etc. While a person can guarantee doctor-patient confidentiality, machines make no such promises.

A malfunction due to algorithmic bias or a failure to maintain the system can result in the loss of this data. Or worse still, it can get into the hands of the wrong people who can easily use this information against the people concerned. This occurs when the system isn’t properly secured against hackers. Indeed, according to a study published by ScienceDaily, advances in artificial intelligence pose a threat to people’s health data. It’s necessary to find and take appropriate measures to secure any AI system in the healthcare industry.

The Road Ahead for Dr. AI

Paving the way for AI to take over the healthcare industry and replace doctors completely isn’t going to be easy. For one, we will have to find ways to address and overcome the above-mentioned challenges. Further studies will need to be conducted for identifying and countering potential risks. Finally, we’ll need to find ways to educate and convince people of the various benefits AI can bring to the healthcare industry so that they feel safe in putting their trust in machines rather than humans. Indeed, according to a survey by Accenture, one-fourth of the patients who claimed that they wouldn’t use AI-powered health services.

Not only do the consumers need to be educated about the advantages of AI, but healthcare professionals should also be informed about the use of algorithms. Only when they’re able to put their trust in these algorithms can they be convinced to use them in their facility. And this trust can only be built through a large amount of clinical validation which can only be achieved by conducting expansive studies and researches in the field.

There can be no doubt that we still have a long way to go before AI can completely take over the healthcare industry since it still has to prove its worth in smaller facilities and in developing countries. There’s no denying the fact; however, those who have already incorporated AI into their facility are enjoying a competitive advantage. Yet, even in such places, AI is facilitating doctors with their diagnosis, etc. instead of attempting to replace them completely.

We need to remember that AI is highly dependable on machine learning algorithms; only a person, a specialized doctor, in this case, can view a patient holistically and consider several other factors before coming up with a treatment plan. Plus, as Eliezer Yudkowsky, the co-founder of the Machine Intelligence Research Institute, warns, “By far, the greatest danger of Artificial Intelligence is that people conclude too early that they understand it”. It’s imperative that professionals have a comprehensive knowledge of the system they’re using and are familiar with ways to secure it. Hence, while the benefits of AI to the healthcare industry are no doubt numerous, AI, according to HitConsultant, should assist, not replace professional doctors in the healthcare industry.

Written by Asim Rais Siddiqui, co-founder and CTO at TekRevol.