Artificial intelligence has the potential to help healthcare systems reach their ‘quadruple aim’ by making connected and AI-enhanced care, precision diagnostics, precision therapeutics, and ultimately, precision medicine more accessible and standardized. Research regarding the use of AI in healthcare is advancing quickly, with various potential applications being showcased throughout the healthcare industry (encompassing both physical and mental health). These applications include drug discovery, virtual clinical consultations, disease diagnosis, prognosis, medication management, and health monitoring.
AI in the present and the near future
AI systems are not reasoning engines at this time; that is, they are unable to reason in the same way as human doctors, who can rely on “clinical intuition and experience” or “common sense.” Twelve AI, on the other hand, works as a signal translator, interpreting patterns in datasets. Healthcare organizations are now starting to use AI technologies to automate repetitive, time-consuming procedures. Additionally, there has been significant advancement in proving the application of AI in precision diagnostics (such as radiotherapy planning and diabetic retinopathy).
Medium-term AI (the next five to ten years)
The creation of strong algorithms that are effective (i.e., require less data to train), able to use unlabeled data, and able to integrate disparate structured and unstructured data—such as imaging, electronic health, multi-omics, behavioural, and pharmacological data—will likely advance significantly in the medium term, according to our proposal. Furthermore, medical practices and healthcare organizations will go from merely implementing AI platforms to working with technology partners to co-innovate new AI systems for precision medicine.
AI in the long run (more than ten years)
AI healthcare systems will eventually reach a state of precision medicine through AI-augmented healthcare and connected care as AI systems grow cleverer. A preventative, individualized, data-driven illness management paradigm that improves patient outcomes (better patient and clinical experiences of care) in a more economical delivery system will replace the existing one-size-fits-all approach to healthcare.
Augmented/connected care
Through the care pathway, AI could improve patient flow and experience, greatly reduce healthcare inefficiencies, and improve caregiver and patient safety. For instance, AI could be used to remotely monitor patients (e.g., intelligent telehealth through wearables/sensors) in order to identify and promptly care for patients who are at risk of deteriorating.
In the long run, we anticipate that passive sensors combined with ambient intelligence will connect healthcare clinics, hospitals, social care agencies, patients, and caregivers to a unified, interoperable digital infrastructure. Two AI applications in linked care are listed below.
Chatbots with artificial intelligence and virtual assistants
In primary care and community settings, patients are using AI chatbots, like those found in Babylon and Ada, to recognize symptoms and suggest additional steps. Wearable technology, such as smartwatches, can be connected with AI chatbots to give patients and caregivers advice into how to improve their behaviour, sleep, and overall health.
Intelligent and ambient care
Additionally, environmental sensing without the need for any peripherals has emerged.
● Faculty and researchers at the Massachusetts Institute of Technology invented Emerald, a wireless, touchless sensor and machine learning platform for remote sleep, breathing, and behaviour monitoring.
● Google Nest claims to use motion and sound sensors to track sleep, including sleep disruptions like coughing.
● An article that was just released examines the possibility of contactless cardiac rhythm monitoring with smart speakers.
● Automating administrative tasks like recording patient visits in electronic health records, streamlining clinical workflow, and freeing up clinicians to spend more time on patient care are all possible with AI systems that use natural language processing (NLP) technology (e.g., Nuance Dragon Ambient experience).
Precision Diagnostics
Diagnostic imaging
The most popular AI application at the moment is the automated classification of medical pictures. More than half (129 (58%) and 126 (53%) of the AI/ML-based medical devices licensed in the USA and Europe between 2015 and 2020 were approved or CE marked for radiological use, according to a recent assessment. In image-based diagnosis, studies have shown that AI can match or surpass human experts in a number of medical specialties, such as radiology (a convolutional neural network trained with labeled frontal chest X-ray images outperformed radiologists in detecting pneumonia), dermatology (a convolutional neural network trained with clinical images was found to classify skin lesions accurately), pathology (one study used whole-slide pathology images to train AI algorithms to identify breast cancer lymph node metastases and compared the results with pathologists’ findings), and cardiology (a deep learning algorithm detected heart attacks with performance on par with cardiologists).
We acknowledge that the NHS has several exemplars in this field, such as the
National Pathology Imaging Co-operative and the University of Leeds Virtual Pathology Project, and we anticipate that AI-based diagnostic imaging will be widely adopted and scaled up in the medium future.
Diabetic retinopathy screening
Diabetic retinopathy screening and timely treatment are essential to lowering avoidable, diabetes-related visual loss globally. However, considering the large number of diabetic patients and the shortage of eye care personnel globally, screening is expensive. 40 Studies conducted in the USA, Singapore, Thailand, and India on automated AI algorithms for diabetic retinopathy have shown good diagnostic performance and economic viability. 41–44 Additionally, the Food and Drug Administration-approved AI algorithm “IDx-DR,” which showed 87% sensitivity and 90% specificity for identifying more-than-mild diabetic retinopathy, was approved for Medicare coverage by the Centres for Medicare & Medicaid Services.
Reducing waiting times and increasing accuracy in radiotherapy
Helping physicians with picture processing and planning duties for radiation cancer treatment is a significant use of AI. At the moment, segmenting the photos is a tedious and time-consuming process that is done by hand by an oncologist who uses software specifically made for this purpose to create contours around the areas of interest. Wait times for beginning potentially life-saving radiation therapy can be significantly decreased thanks to the AI-based Inner Eye open-source solution, which can minimize this preparation time for prostate and head and neck cancer by up to 90%.
