Few industries are ripe for creative disruption than healthcare, which consumes vast sums of private and non-private resources while still struggling to fulfill the demand of a growing and aging population all over the world.
Artificial intelligence is due to this fact increasingly becoming a serious force within the medical field. But like many industries, its full potential is simply now starting to be realized, and uncertainty stays about how best to make use of AI to deliver higher, more efficient care – and improve the experiences of patients and healthcare professionals.
Diagnostics and imaging
Perhaps the world where AI has attracted essentially the most attention is its potential to enhance the speed and accuracy of interpreting diagnostic scans.
For example, Imperial College Health Partners within the UK – which brings together NHS providers, universities and industry in north-west London – sees an enormous role for technology in delivering healthcare innovations based on real-world evidence. Its CEO Axel Heitmüller says that AI systems already make it possible to read MRI and CAT body scans in addition to X-ray images “perhaps more consistently than humans can.”
However, he cautions against leaving clinical professionals out of the equation: “Despite all of the hype, it seems that one of the best results are achieved once you mix humans and machines.”
One area that requires further discussion is the premise against which AI tools in diagnostics needs to be judged, he says. “Everyone at all times focuses on the machine and complains that a machine shouldn’t be perfect. But we’ve never had (perfection) in medical professionals and that begs the query: What is a suitable failure rate for people?”
Pranav Rajpurkar – Assistant Professor of Biomedical Informatics at Harvard Medical School and co-founder of a2z Radiology AIwhich has created an AI model for abdominal and pelvic CAT scans, believes that failure rates could at some point be completely eliminated. He says: “I believe there is usually a world where, because of AI, we don’t make medical errors and where no disease is missed.”
However, although the technology already contributes to the sooner detection of time-critical conditions, it doesn’t yet make people “faster at what they do,” he warns.
His research mission is to develop so-called generalist medical AI models that can have the ability to “perform the total range of tasks that physicians can perform when interpreting medical images.”
For example, while AI can currently detect lung nodules on chest X-rays or lesions on mammograms, “the expert still has to do 200 to 400 other things as a part of the interpretation, and we don't have algorithms (yet) that may do this,” emphasizes Rajpurkar.
“The efficiency value proposition is one which we’ve got not yet implemented in AI, but I believe we’re on the verge of having the ability to implement it with upcoming technological advances.”
Treatment
Another area where AI has already proven itself is within the delivery of tailored treatments. Anna Sala, an allergist who leads the innovation unit on the Vall d'Hebron Barcelona Hospital Campus, points to a European project called TRUSTroke that goals to optimize stroke treatment. The project recently accomplished a successful pilot coordinated within the Spanish hospital.
The system is trained using data from medical records and other information provided by patients and healthcare professionals via a mobile app. Sala says the platform will “analyze all aspects related to the pathology, the patient and their environment.”
The resulting information “will provide doctors, patients and caregivers with reliable guidance to personalize treatment as much as possible and stop risks and complications.”
She adds that AI has also began to indicate its prowess in the world of rare diseases, pointing to a platform that has arrived at diagnoses that even doctors couldn't make after having details about symptoms and medical history of the patients had received.
Communication with patients
At the identical time, AI helps make interactions between healthcare teams and patients easier and more productive. For example, it could actually be used to power a tool that transcribes a patient's consultation and allows the doctor to keep up eye contact, protected within the knowledge that an account can be created that may be quickly shared with the patient.
Even more ambitious, AI helps determine when patients need a follow-up visit. Sala cites a chatbot called Lola, also developed in Vall d'Hebron with two firms, AstraZeneca and Tucuvi. It provides personalized follow-up to patients with heart failure and chronic obstructive pulmonary disease by asking those that have been hospitalized to reply a series of questions on their phone about how they’re feeling and their ability to perform on a regular basis tasks.
Your answers are sent to the cloud for evaluation by AI. If they cause concern, patients are asked to make an appointment. This innovation saves patients unnecessary trips and, in keeping with Sala, also protects the environment by reducing the carbon footprint.
Back office
AI's ability to enhance administration may not grab headlines, but it could actually be just as transformative as more noticeable developments in patient assessment and care, says Heitmüller.
Most industries haven't began with customer-focused innovation yet, he points out – pointing to the advocacy community that has used AI “to automate really boring, repetitive processes like database searches, etc.” But in healthcare, we at all times appear to have a conversation to begin with the query: 'Should we’ve got an AI doctor?' relatively than 'Have we actually automated our back office?'”
At least within the UK public health system (NHS), there may be little incentive to enhance this aspect of operations, he admits. “We have annual budgeting, so if you happen to are successful, any savings can be taken away from you. . . and due to this fact it doesn't make much difference whether AI has something to supply.” But integrated care systems within the USA, for instance, may benefit from the technology, he suggests.
However, investment stays an obstacle. Despite the “great potential for automation,” “this area shouldn’t be where the funding is, neither is it where the eye is” in healthcare.