AI’s huge potential to improve healthcare

Artificial intelligence is increasingly part of life, and so are anxieties about how it will change life as we know it. How it will change our jobs is just one aspect of the dystopian future we imagine it is creating. Some, if not many, of these concerns warrant serious consideration, but you don’t need to be wearing rose-tinted glasses to see that AI also holds immense potential to improve our lives, particularly in healthcare.

AI’s powerful image-processing capabilities are already helping healthcare professionals to detect serious diseases early by accurately analysing medical images. For example, AI algorithms significantly help radiologists spot malignant tumours, and in emergency rooms, some medical facilities use AI to spot embolisms and signs of stroke, which is crucial for minimising the risk of brain damage.

AI is also being used to predict surgical outcomes and the best course of patient treatment by analysing the patient’s medical history, genetics, and current condition. Surgeons use AI-powered surgical robots to conduct complex and precise surgeries with minimal invasiveness, leading to faster recovery times, reduced complications, and improved patient outcomes. AI is revolutionising many aspects of health provision and delivery, such as diagnostics and drug discovery.

It’s worth noting that AI is not one technology but an umbrella term for several technologies. Currently, it is dominated by machine learning algorithms via which we can build computer software that learns to perform specific tasks independently, instead of providing detailed instructions to software on how to perform the tasks. Deep learning is also an umbrella term referring to a subset of machine-learning algorithms that try to mimic human learning processes via mathematical constructs known as artificial neural networks.

My career as a software engineer began two decades ago when I focused on using AI to streamline software production. However, a growing fascination with its potential to improve lives beyond software development led me to concentrate on health AI. I am a deep learning software engineer, and my passion lies in leveraging AI for a greater purpose: building AI software tools to empower individuals with disabilities and better equip healthcare professionals to support them.

Among my current projects is using AI to identify autistic children early. Autism is a lifelong developmental condition affecting individuals’ social and communication skills. About 1-2.7 percent of children are autistic. The lifetime cost of supporting someone with autism in the US is estimated to cost US$1.4-2.4 million. In New Zealand, 27 percent of autistic kids are not identified by the age of six or seven – three years later than optimal for delivering early language and other key supports.
 

This means it can be too late for a quarter of our autistic kids to benefit properly from the support available. This delay could be for many reasons, but mostly because families may not seek support in time, and when they do, the process could be very time-consuming and costly because of its complexity. Often, multiple autism specialists need to be involved to confirm diagnosis. If families live in remote areas, access to one specialist could be limited, let alone several. Given the global shortage and competition for healthcare workers, this isn’t likely to change in the near future.

To address these challenges, researchers introduced autism screening techniques that are mostly easy to administer and affordable but also often inaccurate. We are building an AI platform called Autism Artificial Intelligence which learns autistic behavioural indicators to automatically assess toddlers for autism as early as 18 months old. We aim to make Autism AI accessible and easy to use using autism screening methods with a level of accuracy close to medical diagnosis.

Another area of my research focus is the development of AI that understands atypical speech, speech that has been affected by a stroke, an accident, or a disease that leads to motor speech disorder. AI can bridge this communication gap by allowing you to interact with your phone using what many would find to be unintelligible speech.

The phone translates it for others, talking to them on your behalf in intelligible speech or even in other languages. Making computers understand atypical speech also enables us to unlock new tools and technologies, such as designing an automated speech therapy and monitoring system to help people with speech disorders find their speech in the comfort of their own homes, wherever they are.

Using AI to identify signs of dementia early is another area of health AI I’m investigating. We’re developing AI models to detect early signs of dementia by analysing speech patterns to find memory flaws and other dementia symptoms. There’s no cure for dementia yet but identifying it as early as possible allows us to manage and possibly slow its progress. Such technology could provide a crucial 24/7 screening tool, allowing us to seek intervention sooner and ensure we spend quality time with our loved ones. This project is personal for me as my mother has been diagnosed with Alzheimer’s disease. I’m witnessing her decline and wishing I’d started this project earlier.

These are just a few examples of how we use AI to transform healthcare services to be more accessible and affordable, to relieve the pressure on our precious medical doctors and nurses and give them mechanisms to support us better. This future is not science fiction; it’s the exciting potential of AI in healthcare waiting to be realised. With continued investment and responsible development, AI can revolutionise how we deliver and experience healthcare, saving lives and improving wellbeing for all.