The proliferation of Artificial Intelligence and Machine Learning in Healthcare Pt. 1
We are not even a quarter-way through the century, and what seems to be one of the most revolutionary emergent technologies and facilitators for human society of this century is poised to be Artificial Intelligence (AI). It is a well known notion that artificial intelligence and related services and platforms are poised to revolutionize international productivity, working patterns, and lifestyles and generate massive wealth.
For instance, McKinsey sees it furnishing international economic activity of approximately $13 trillion in 9-years time from now. In the immediate term, Gartner predicts that worldwide AI-driven economic activity will appreciate from approximately $1 trillion in 2018 to approximately $4 trillion in a year’s time from now.
It is well known that this revolution has been incited by, to a dominating extent, driven by the potent machine learning (ML) utilities and strategies like Deep Convolutional Networks, Generative Adversarial Networks (GAN), Gradient-boosted-tree (GBM), Deep Reinforcement Learning (DRL), etc.
But, conventional enterprises and technology domains are the not the sole areas being affected by artificial intelligence. Healthcare is a domain that is believed to be very apt for the applications of AI utilities and strategies.
Mandated practices like Electronic Medical Records (EMR) have already prepped healthcare systems for application of Big Data utilities for next-gen data analytics. Artificial intelligence/ML are pre-destined to inject even more value to this flow. They are forecasted to improve the quality of automated technologies and smart decision-making in primary and tertiary healthcare and public medical frameworks. This could be the largest effect of artificial intelligence tools as it can possibly revolutionize the quality of life for billions of people around the world.
Critical instances the application of ML in medicine
AI-assisted radiology and pathology
Currently, electronically recorded medical imagery data is abundant and DL algorithms can be inputted with the variant of dataset, to identify and find out patterns and anomalies. Machines and algorithms can undertake interpretation of the imagery data very much like a well trained radiologist would – detecting suspect spots on the surface of the skin, tumours, lesions, and bleeding in the brain. The harnessing of ML utilities / Artificial Intelligence / Platforms for assisting radiologists is hence, poised to grow by orders of magnitude.
The strategy furnishes a solution to a crucial problem in the domain of medicine as, throughout the globe, well-versed radiologists are becoming a hard catch. They’re in high value. In a majority of scenarios, these skilled employees are under massive pressure owing to the overload of digital medical information. An average radiologist, is required to generate interpretation outcomes for a single image each 3—4 seconds to fulfil the demand.
Detecting uncommon or though to diagnose illnesses. This is often dependent on identification of the so-named “edge-cases”. As this type of ML framework is developed on massive dataset consisting of raw imagery (and several transformations) of these illnesses, they are usually more reliant than humans for this variant of detection.
They are predicted to improve the quality of automation and smart decision-making in primary/tertiary patient care and public healthcare frameworks. This could be the largest impact of artificial intelligence utilities as it can possibly transform the quality of life for the global populace.
An excellent evaluation instance is Microsoft’s pet project InnerEye which deploys Machine Learning techniques to segment and detect tumours harnessing 3D radiological imagery. It can assist in precise surgery planning, navigating, and efficient tumour-contouring in relation to radiotherapy planning.
MRI and other sophisticated imagery systems, increasingly leveraged for early cancer detection, are being outfitted with Machine Learning algorithms.
ML utilities are also injecting considerable value through augmentation of the surgeon’s display with data like cancer localization during robotic procedures and other image-directed interventions.
The utilization of machine learning / artificial intelligence utilities for assistance of radiologists is, thus, poised to grow exponentially owing to innovation.
Data Science and Machine Learning for drawing actionable insights
In the current scenario, exabyte-sized medical information is presently undergoing the process of digitization at several healthcare institutions (public hospitals, nursing homes, doctor’s clinics, pathology laboratories, etc.) Unluckily, this information is typically messy and unstructured. Differing from typical transactional variations of business information, patient information is not amenable to simplistic statistical modelling and analytics.
Robust and AI-driven platforms, usher in innovation – with the capacity of connecting with an array of patient databases and to undertake analysis of a complicated mixture of data types (for example, genomics, pathology, radiology imagery, medical history) are the requirement of the hour. Further, these frameworks should have the capacity to sift through the analyses in an in-depth fashion and find out the hidden patterns.
Also, they should possess the means to translate and visualize whatever they have found out to forms comprehensible to humans so that medical practitioners and other medical pros can work on their output with increased confidence and complete transparency.
Interpretable Artificial Intelligence and distributed Machine Learning systems are adequate for these purposes and are set to fill the requirements for these frameworks in the immediate future.
Physical bots for surgery assistance
Surgical bots can furnish one-of-a-kind assistance to human surgeons,
- Improving the capacity to observe and navigate in a process/procedure
- Developing precise and minimally invasive incisions.
- Causing reduced pain with optimal stitching geometry and wound
There are really exciting prospects for the harnessing of Machine Learning / Artificial Intelligence for such digital surgery bots.
- A software-driven collaboration of bots with the assistance of massive distributed processing
- Data-driven insights and direction on the basis of surgical histories (carried out by human beings and machines alike) and their results (favourable or not)
- Artificial intelligence-produced virtual reality space for real-time direction and guidance
- Potential for telemedical and remote surgery for comparatively simple processes
Artificial Intelligence – It’s role in healthcare operations management and patient experience
Particularly within the United States of America, the expense and complications in relation to obtaining adequate health care, by the common public, has been a topic of protracted and bitter debate.
Artificial Intelligence and related data-based strategies are uniquely positioned to handle some of the issues, detected as the root reasons – long queues, fear of inflated bills, the long-drawn and overtly complicated appointment procedures, not obtaining access to the correct healthcare practitioners.
Those same sets of issues have been impacting conventional enterprises for several decades and Machine Learning / Artificial Intelligence strategies are already an aspect of the solution. This is owing to the fact that massive databases and smart search algorithms, which are the domain of artificial intelligence frameworks excel at these pattern matching or optimization problems. Thus, sophisticated Artificial Intelligence / Machine Learning utilities ought to be harnessed by hospitals and public health organizations in their daily operations aspects.
The amazing thing is that the worry with regards to information privacy, which is a complicated and tough problem for healthcare frameworks, does not put forth a massive challenge to this variant of application of Artificial Intelligence. Most typically, an operational issue does not consist of confidential patient information in relation to illness, diagnoses, or medications, however, much like any other advanced business enterprise, is made up of data connected to capital, finance, marketing or HR issues.
The core objective for these frameworks should be to make the artificial intelligence-driven platforms aiming to improve the experience of healthcare services for the biggest subsection of the common populace. The overarching objective for the biggest section of typical people. The overarching objective of already-deployed frameworks in conventional enterprises is to maximize profit. Potent artificial intelligence utilities concerned with healthcare operations-management must distinguish themselves from those traditional frameworks by combining empathy with the goal of merely padding out the bottom line.
Creation of new drugs with the assistance of Machine Learning / Artificial Intelligence strategies
Machine Learning and Artificial Intelligence are increasingly being selected by major names in BigPharma to identify solutions to find solutions to the devilishly complex issue of successful drug discovery. A few prominent instances are Sanofi, Genentech, and Pfizer. All types of therapeutic domains – metabolic illnesses, cancer treatments, immune-oncology drugs are spoken of in various case studies that you can find with a quick Google Search.
Moving beyond the traditional long-haul procedure, Artificial Intelligence strategies are more and more being harnessed to accelerate the basic processes of early-stage candidate selection and mechanism discovery.
For example, biotechnology enterprise Berg leverages its artificial intelligence platform to undertake analysis of massive amounts of biological and outcomes data (metabolite, lipid, enzyme, and protein profiles.) from patients to illustrate critical variations amongst diseased and healthy cells and detect unique cancer mechanisms.
Another noteworthy instance in this regard came from DeepMind’s publication of the potential protein structures connected with the COVID-19 virus, (SARS-CoV-2) leveraging their AlphaFold system.
Several start-up organizations are also working on leveraging AI-frameworks to undertake analysis of multi-channel data (patents, research papers, clinical trials, and patient records) through harnessing the latest strategies in Bayesian inference, Markov chain models, RL and NLP. Detection of patterns and production of high-dimensional representations, to be documented in the cloud and leveraged in the drug-discovery procedure, are the vital goals.
Robust artificial intelligence utilities for healthcare operations-management ought to distinguish themselves from those typical frameworks by bringing together empathy with the goal of profit production.