By Sam Rydberg-Cox, Spring 2021.

As the global COVID-19 pandemic consumed every aspect of our lives, it also catalyzed the expansion of technological advances. Telehealth has exploded, and if you follow the money, it seems as though these advances are here to stay. Compared to the beginning of last year, investment in telehealth has nearly doubled to $4.2 billion. [1] Investments in artificial intelligence have also exploded, ranging from $20 million to $200 million, with companies using AI and machine learning to create better patient interactions on the phone, streamline revenue management, and even predict kidney health. [1] Overall, in the first quarter of 2021, another $9 billion has been funneled into AI companies. [1]

How do these AI companies, in particular, justify receiving such large sums of money? As mentioned above, their services range from patient interaction to money management to diagnosis. At their core, these companies streamline processes that typically require huge amounts of manpower and time to complete. As a result, their advancements are allowing hospital staff to provide better care by reducing their workload and by directly assisting in patient evaluations. A good example is Strive Health, which focuses on improving kidney care through predictive analytics and machine learning in order to form personalized treatment plans for patients. The company partners with healthcare providers to improve patient experience and lower patient costs. [2]

Undoubtedly, the most important aspect of these technologies is their potential to improve patient care. For another example, a research team at MIT that focuses on skin cancers, specifically melanoma, took a machine learning approach to try and improve detection mechanisms for this cancer. Normally, the diagnosis process for melanoma includes a visual inspection by a dermatologist to identify potential malignant lesions. To improve this, these researchers have developed artificial intelligence that can analyze images from a smartphone and determine the risk of a lesion becoming cancerous. [3] The system uses deep convolutional neural networks (DCNNs) to analyze the images. [3] The DCNNs are able to classify the pictures into certain categories– similar to the technology in your phone that tries to organize your photos by who is in them. In the case of this mechanism, the groups are not separated by the faces of your friends and family but by the risk level that a certain lesion becomes cancerous. This allows patients to use their cell phone camera to potentially identify potentially cancerous lesions and either dismiss them as safe or be alerted that it needs to be further checked out. This makes it easier for patients to identify potential cancer earlier and in turn increase survival rates. When compared with the opinions of dermatologists, the AI was more than 90% successful, and in a much shorter period of time. [3]

So why haven’t these technologies been expeditiously introduced into the medical field yet? As the field is fairly new, it still has some issues to iron out, and the implementation of this technology is fairly difficult. There are two main reasons for this. Firstly, healthcare is a vastly complex field influenced by a number of political, economic, and moral factors. The addition of AI technology by itself is not enough to make it an integral part of the healthcare system and fulfill its full potential, as the technology must be successfully integrated and appropriately monitored. [4] Secondly, it is very hard to account for biases that may arise. [4] In the example of melanoma detection, the samples used to train the system were mostly from Madrid. [3] This means that the system may have trouble detecting cancerous lesions when it receives a picture of someone whose skin color is not prevalent in Madrid. Further, other biases can arise within the training sessions of the AI that must be addressed.

Action to remedy these issues is in place, but the field is still evolving. A pivotal step forward is compiling databases that can be used for training these machines. There are obvious challenges that come along with gathering pictures of a patient’s condition or other information, but an accessible database would be a large step forward. These databases would allow for easy access to large datasets to train the AI and could potentially help to reduce bias if it is made to include a sample representative of the population. Some hospitals have begun this process and it has shown great promise in terms of advancing the technology. As their development becomes more widespread, medical AI will be able to advance the field of medicine even further.  

1. Heather Landi. 2021. “Global Investment in Telehealth, Artificial Intelligence Hits a New High in Q1 2021.” FierceHealthcare. April 20, 2021. https://www.fiercehealthcare.com/tech/global-investments-telehealth-ai-startups-reached-record-levels-q1-2021.
2. “Strive Health Raises $140 Million Led by Alphabet’s CapitalG to Tackle $410 Billion of Unmanaged Kidney Disease Spend.” 2021. March 16, 2021. https://www.businesswire.com/news/home/20210316005527/en/Strive-Health-Raises-140-Million-Led-by-Alphabet%E2%80%99s-CapitalG-to-Tackle-410-Billion-of-Unmanaged-Kidney-Disease-Spend.
3. “An Artificial Intelligence Tool That Can Help Detect Melanoma.” n.d. MIT News | Massachusetts Institute of Technology. Accessed April 21, 2021. https://news.mit.edu/2021/artificial-intelligence-tool-can-help-detect-melanoma-0402.
4. Panch, Trishan, Heather Mattie, and Leo Anthony Celi. 2019. “The ‘Inconvenient Truth’ about AI in Healthcare.” Npj Digital Medicine 2 (1): 1–3. https://doi.org/10.1038/s41746-019-0155-4.