Machine Learning in Healthcare and Medicine
By Vishnu Dasu
“You can have data without information, but you cannot have information without data”
– Daniel Keys Moran.
Current generation robotics and computer technology are far from replacing doctors and nurses. However, Machine Learning and AI are, without a doubt, transforming the healthcare industry. These powerful tools could aid physicians in making better decisions and decrease the chances of overlooking potentially life threatening diseases.
Machine learning is defined as a data analysis approach that automates analytical model building. Ingenious algorithms iterate over the supplied data and locate (otherwise untraceable) information, without being explicitly programmed to do so. This iterative aspect of machine learning enables computers to “learn from previous computations and patterns to produce reliable, repeatable decisions and results” according to LA-based RX4 group, which provides business support to health-care providers.
Machine Learning and statistical methods have been around for a long time, dating back to the early fifties. However, it is only recently that they have become feasible to implement on large data so as to produce effective results, credited to the recent advances in computer processing technologies.
The number of companies investing their resources in AI monitored healthcare shows that it is an area with a promising future. IBM, Lumiata, Google, Nervanasys, and Sentient.ai are some of them. Frost & Sullivan, a business consulting firm, predicts that artificial intelligence in healthcare will see a “dramatic market expansion” in the years to come, with the potential to reduce the cost of medical treatments by half. The impact of ML in the healthcare industry is wide and varied.
Managing Medical Records
This is probably one of the most obvious applications of ML in the healthcare industry. The medical history of a patient can provide valuable insight, aiding physicians in predicting future diseases and complications that could deter the patient’s health. Analysing the medical history of the patient’s antecedents could help predict and maybe even impede congenital disorders. Recently, DeepMind – Google’s subsidiary that built AlphaGo, launched the Google DeepMind Health Project which is used to mine data of medical records to provide faster and effective health care services. Still in its initial stages, DeepMind is currently cooperating with the United Kingdom’s NHS Foundation Trust in the field of ophthalmology.
ML algorithms are proficient at recognizing patterns in data, which is what diagnosis is all about. Lumiata, the AI-powered predictive analysis company, is using ScienceOps to incorporate health risk algorithms into their predictive tool called the Risk Matrix. These predictive analytics tools can provide accurate insights and make predictions related to symptoms, diagnosis, and medications for patients. Another company called Pathway Genomics – which is backed by IBM – is developing a simple blood test to help in the prognosis of certain cancers. Microsoft too is exploring the healthcare industry with its recent initiative known as InnerEye. The team is currently working on image diagnostic tools for image analysis. This could help identify abnormalities which could otherwise be easily overlooked by the naked eye.
As a matter of fact, Dr Bhargav Bhatkalkar (Assistant Professor in the Dept. Of CSE, MIT) is working in collaboration with Kasturba Medical College on a project that aims to predict macular degeneration so as to facilitate treatment from an early stage.
Treatment and Suggestions
Diagnosis is a very complicated process and involves a number of factors which machines presently cannot comprehend. However, computers can aid physicians in making the right decisions regarding diagnosis and treatment.
IBM has been working alongside Memorial Sloan Kettering’s (MSK) Oncology department to suggest treatment ideas and options to doctors dealing with cancer cases, with the help of IBM’s supercomputer (Watson) and MSK’s massive corpus of medical data amassed over the years. In the field of psychology, Ginger.io is developing an app to remotely deliver mental health treatments. This apps allows users to analyse their own moods over time and cope using strategies that have been developed by doctors.
Crowd sourced Medical Data Collection
Current health data is invaluable as it can help in the prognosis of chronic diseases and provide treatment at the initial stages of the disease. IBM, which is bending over backwards to acquire all the data it can, has partnered with Medtronic to make sense of diabetes and insulin data in real time and has also bought Truven Health, a healthcare analytics company, for $2.6 billion dollars. Apple’s ResearchKit aims to collect and analyse data to aid in the treatment of Asperger’s syndrome as well as Parkinson’s disease by allowing users to assess their conditions over time, which in turn feeds the ongoing progress data into an anonymous pool for future study.
Robotics in surgery
As mentioned before, we are still long way from developing intelligent systems that can completely eliminate human interaction in performing complicated surgeries. However, robots have been developed which could aid surgeons in intricate and arduous operations.
The da Vinci Surgical System is a robotic surgical system developed by Intuitive Surgical. This device allows surgeons to manipulate its dexterous limbs to perform surgeries with finer detail in tight spaces where it would otherwise be a herculean task for a human hand to perform alone. This also facilitates a reduction in the number of incisions being made which would reduce the recovery time for the patient.
Given the advancements that have taken place in the health industry, it is safe to say that machine learning has more than simply scratched the surface. However, we are far from being completely reliant on machines. There are quite a few drawbacks associated with implementing machine learning techniques. Machine learning algorithms have some ambiguity associated with them. In other words, these algorithms do produce a result but are unable to explain why and how they produced that result. Not being able to warrant the result produced means that physicians cannot completely depend on computers, yet, for decision making.
Another drawback is that a computer treats the data received from medical records just as it would, any other data. Though analysing medical records can give a fair idea of the diagnosis and treatment to be followed, a doctor’s involvement is still necessary. Finally, it is also important for patients to get accustomed to artificial intelligence. It is important to dismiss the fears and prejudices that we have against artificial intelligence.
The challenges faced are simply a motivation for further innovation. Computer scientists are still improving healthcare and medicine, and people are gaining awareness of the benefits of artificial intelligence.