An unspoken promise of all technological progress is that the average citizen's life will improve greatly from its current level. The improvement of life quality is an explicit promise of the current AI summer, and in particular machine learning has promised to improve and lower the cost of medical care. Companies such as Babylon Health have effectively promised the patient a low cost personal physician. Over the course of this blog post we will speculate about the near and immediate future based upon the current research being conducted in academia and industry in the area of medicine and wellness.
We know that our health is dependent upon our diet. Research has shown that a balanced diet and moderate exercise is a protection against a number of late life diseases such as cancer and dementia. However with busy lifestyles where food preparation speed is favoured over good quality food, and misleading food labelling makes it difficult to have a balanced diet. There is now a movement with the machine learning community to develop food recommendation systems that actively recommend food that protects against specific diseases . It is not a far fetched idea that personal digital assistants on mobile devices can produce personal diet plans which maximize the health of the user. And with automated ordering it would be possible for them to order the food and monitor the health outcome through sensors embedded in the household.
The food as medicine is being supported by another area of AI, knowledge representation also known as semantics. Pressure from organizations such as the USDA for better nutritional information has pushed scientists to produce knowledge graphs or ontologies of food. These ontologies describe food and food products in a standard way, and therefore can be used in machine learning algorithms and in the near future automated cooking systems such as Thermomix. The start of this revolution can be seen on the bbc website where recipes are published in JSON-LD.
Food as medicine is a cheap and effective way of combating disease, however AI is not limited to the mundane ordering of food, and is being used as a diagnostic tool.
Mobile phones can be at the centre of personal health care. It is not a massive leap of the imagination to visualize a personal AI physician sitting on a mobile phone, and asking patients to submit photos of various parts of their bodies so that diseases can be detected. It is possible to predict diseases from images such as Alzheimer's disease and skin cancer therefore the AI physician could suggest the patient submit photos based upon their lifestyle or risk profile, such as if a patient has booked multiple sessions on sunbeds or has been exposed to excess Ultra Violet. The risk could be estimated by the AI physician from sensors on the phone, GPS data and Internet purchase history. This preemptive scanning and early diagnosis would prompt early medical intervention which always has better outcomes than later diagnosis.
The ultimate goal of an automated AI physician would be not to detect disease, but to recommend treatment. And if possible automatically order and supervise the consumption of medication. And again with current technology it is not a huge leap to imagine a fully automated system on a mobile device that is constantly monitoring the user's health and intervening when necessary.
The imagined utopia of the near future is not limited to a personal physician. Cures for the major diseases may be here sooner than we can imagine. It is not a wild claim to state that we will be entering a golden age of drug discovery which is being driven by deep learning. Traditionally new drugs were discovered through researcher intuition and long laborious tests in wet labs. Discovery of new drugs therefore is a decade long process. Healthcare AI and deep learning has presented a shortcut where large numbers of drug candidates can be simulated in a computer and tested for their theoretical properties and the most promising candidates are then handed off to the wet lab researchers for evaluation. This process takes luck and time out of the equation in the drug discovery process. Therefore in our imagined health utopia that there will be drugs designed and discovered by AI that will offer relief and possible cures for the major diseases. This is not a forlorn hope, and already there drug candidates such as BPM31510, an anti-cancer candidate which was discovered by an AI process, is in phase two trials. In our imagined health utopia there will be possible treatments for diseases which are at the moment nothing short of a death sentence.
Technology often offers false promises, but this is not the case with AI and medicine. AI can help with disease prevention, detection, treatment as well as drug discovery. These techniques applied at a large scale will have a significant impact on a population' s health. It is possible to imagine that AI assisted medicine can not only extend people's life expectancy, but their quality of life as well. The only roadblock on the horizon is that of ethics and privacy.
Invasive AI personal physicians have the potential for abuse. For example will insurance companies demand access to the records and inferences made by these pocket doctors to deny insurance cover? And there is the human factor of accepting life changing decisions from technology. No doubt mistakes will be exaggerated as we have seen from the press around accidents involving self-driving cars. Is this imagined health utopia just around the corner? It is not, but the trajectory that current deep learning and machine learning is on, is not that far away, and the promise of automated medical agents as promised by science fiction may start to be realized during the lifetimes of the current machine learning practitioners.
If you would like some help to find out what AI can do to help you and to find out what we have already built in the field of AI healthcare, have a look at our AI consulting services.