Digital technology first set its footprint in the healthcare space by introducing fitness trackers and nutrition apps, which was followed by more advanced medical apps aiding in tasks like analyzing patients’ symptoms or suggesting physicians’ appointments. Now digital tech promises to solve many more challenging healthcare problems, including ever-increasing costs and the difficulties of patient-specific therapies.
Emerging digital healthcare technologies such as care management software and blockchain are already starting to transform the healthcare business by changing the way electronic health records (EHR) and clinical trial data are handled. Concomitantly, investment in this sector has reached record-breaking levels. Indeed, digital health startups raised $6.8 billion USD in 2018, outnumbering 2017’s figure by $1.1 billion.
Artificial Intelligence – The new flag-bearer of digital healthcare:
Forbes recently surveyed companies around the globe to rank the most influential digital health technologies of 2019. The report found that 24.5% of all companies believe healthcare will be profoundly influenced by artificial intelligence (AI). Furthermore, a recent market research report also supported this prediction, stating that the AI market in healthcare will reach $6.16 billion by 2022.
We all witnessed how IBM’s Watson supercomputer rose to fame as a Jeopardy champion in 2011. This early success has been followed by commercial applications, including automation of insurance analysis and use of IBM Watson in the immunotherapy and drug discovery spaces — although, however, IBM recently discontinued development due to poor commercial performance, a warning sign of the many challenges still to be addressed for successful AI deployment in this space. Other major players are also exploring uses of AI; for example, Philips recently announced their plans to pursue eICU as an AI-powered solution to handle health records in hospitals and to predict advanced treatment decisions.
In addition to the efforts of giants like IBM and Philips, numerous small healthcare startups are also making major progress in applying AI in healthcare ventures. For example, a Boston-based startup named Buoy Health, who is developing an AI-based diagnosis app, made it to CNBC’s Upstart 100 list in 2018, which chose the most promising startups from more than 500 nominees from nine different countries. Like Buoy Health, a number of healthcare startups are now employing AI-based predictive analysis to better address medical problems.
AI to predict patient responses to personalized therapy:
Freenome, a San Francisco–based startup, is taking the early prediction of cancer to a new level through noninvasive blood tests and AI-powered genomic analysis. Freenome trains its AI on thousands of oncogenic DNA biomarker patterns and then detects similar profiles in a patient’s blood sample to enable detection at the earliest possible stages of cancer. In April 2018, Freenome formed a collaboration with leading proteomics company Biognosys to add protein quantification to its cancer detection approach and to help suggest possible precision therapies. QIAGEN also announced a collaboration with Freenome in May 2018, with the goal of developing novel diagnostic tests to predict a patient’s response to a given immunotherapy. Freenome has already shown promising results to predict colorectal cancer through a blood-based test, using its AI platform and whole genome sequencing of the circulating tumor DNA in blood samples. Their next step is to broaden this approach for other cancer types.
AI as a tricorder – Diagnosing illness at your home:
A Boston-based startup called Basil Leaf Technologies has come up with a consumer-friendly AI-driven technology called DxtER, which promises to allow disease diagnosis in one’s own home with zero clinical training. It can predict and monitor over 13 medical conditions (for example, diabetes and sleep apnea) through multiple noninvasive sensors that gather information on a person’s symptoms and physiological functions and decisively match them with his or her personal and family medical history. DxtER was originally developed for and won the worldwide $10 million Qualcomm Tricorder XPRIZE competition in 2017. Following the XPRIZE success, the company received a $2.6 million grant for DxtER’s commercialization.
However, in reality, the first commercially available device applying this technology will likely address one disease state instead of 13, as Basil Leaf Technologies head of user experience Phil Charron reasons that a device predicting 13 medical conditions may take an abnormally long time to receive FDA approval. This has prompted the company to focus on a similar consumer-friendly device dedicated to monitoring congestive cardiac failure. Nonetheless, the company is continuing to commercialize a blood glucose and hemoglobin monitor as well, which is now undergoing Phase I clinical trials and promises to predict a multitude of health conditions.
Early diagnosis based on clinical data records:
A key healthcare challenge at present is the difficulty of access to the huge amount of available clinical data. Proper access and sharing of these data could solve many problems, such as identifying patients with rare diseases who have similar genetic profiles across multiple centers and improving medical decisions by a comparative analysis of the treatments provided. The New York–based startup Prognos is challenging these data silos. Prognos collaborates with diagnostic labs and gets access to their large clinical datasets, then leverages the power of AI to match a particular patient’s clinical profile with the details of medical insurance plans, with which they predict the probability of that individual developing certain diseases in the future.
Investors like Cigna and Merck Global Health Innovation Fund invested a total of $42 million in Prognos to continue its mission. However, accessing clinical datasets can be potentially risky if it does not comply with patient privacy. According to Forbes, there could be two ways of tackling this issue. First, the clinics could keep their own datasets but also share them in a secure “federated” pool and the search engines would extract only a relevant and limited amount of it. Second, patients could sign a withdrawable consent to share their clinical records in “data-inhaling clinics” and would benefit from the data sharing of others as well. It will be interesting to witness whether Prognos follows either of these paths in the future or deals with this issue in a different way.
Conclusions:
Despite the vast promise of artificial intelligence that we have touched upon in these three examples, AI’s predictive power still has limitations. Its technological foundation is built on previously available datasets, which may make erroneous decisions in novel scenarios or with unorganized EHR. A much bigger challenge is to formulate the necessary ethical and regulatory guidelines, for example, for when AI generates false predictions for patients. Nonetheless, we are continuing to make rapid advances toward using AI in the healthcare space, with early successful applications in medical imaging being matched by huge investment. Exciting uses of AI to provide early disease diagnosis and optimize treatment plans are emerging, and now it’s just a matter of time until we will see the predictive power of AI being used to improve our health in unprecedented ways.
