The future of robotics in healthcare and wellness
New products and services related to robotics will continue to enter the existing market and create new markets in the healthcare and wellness sector. For example, Deloitte Access Economics and the Australian Computer Society recently considered how robotics and remote systems can alleviate health workforce shortages and enhance service provision in regional areas [ADP17].
The application of computer vision to medicine is being revolutionised by deep learning, a special branch of machine learning. It relies on ‘so called’ artificial deep neural networks to extract meaningful patterns in training data. This is then used to learn to detect and recognise objects in images, understand spoken language, or control robots. Deep learning has the potential to completely revolutionise the way doctors diagnose and treat diseases in the future.
A crucial prerequisite to train deep neural networks is the availability of training data. More data, and supplementary diverse data, leads to better outcomes. At the moment, the availability of anonymised pathological data (such as CT or MRI scans and radiographs) is very limited, and controlled by hospitals and individual medical practitioners. Access to such anonymised medical data can progress global advances in the detection, diagnosis, treatment, and management of diseases. This would allow remote, and early, diagnosis of conditions such as skin cancer, macular degeneration or diabetic retinopathy. Such advancement could have a major impact in reducing future medical costs. By working towards greater availability and collaborative use of anonymised patients’ tests and histories to build deep learning powered smart devices, Australia can become a global leader in this area.
Due to the vast remoteness of Australia, there is not always a constant local supply of medication for acute and chronic conditions, especially if access requires consultation with a pharmacist. There is a clear need for a modern, transparent and sustainable solution. Remote pharmacies can fill this gap as well as provide a solution to certain emergency situations, such as to aid in disaster response and to support the Defence sector (see Case Study p. 66). The Australian Productivity Commission advocates robotic dispensing of routine medicine and a commensurate reduction in trained pharmacists. A similar opportunity exists for remote medical imaging, where patients are scanned remotely and diagnosed online by a combination of trained clinicians and artificial intelligence (see Case Study p. 68).
Australia has research strengths in biomechanical modelling and the development of simulation and training for both medical practitioners and for training robotic systems before deployment in real-life situations. Transfer learning is the transfer of control knowledge from one robotic system to another. When the source domain is a simulation, transfer learning is less costly and more practical than using a real robotic system in terms of energy, risk, wear and tear, and patient or operator safety, and has the added advantage of being able to be used multiple times. Simulation environments and transfer learning are important in any field where large or expensive machinery is used, and needs to be thoroughly tested.
The future includes the availability of AI-powered smart mobile devices that are trained on available medical data, reducing the pressure on clinical and outpatient services, freeing up doctors’ time to attend and treat more critical cases. It includes logistics robots performing manual tasks such as linen and food distribution in the back-of-house, and greater use of social robots in front-of-house concierge roles, providing public health education, companionship to patients, and assisting medical professionals to do their jobs effectively.