Last Month in Healthcare AI (June 2024)

Contents: ?? 1) Prediction & Diagnosis | ??2) Treatment & Care | ?? 3) Latest Approvals | ?? 4) Generative AI | ??5) AI Assistants | ??6) Governance & Ethics

?? 1) Prediction & Diagnosis

?? AI tool predicts autoimmune disease risk using genomic data. The artificial intelligence-driven EXPRESSO model uses genomic data, epigenetics and other information to identify risk genes for autoimmune disease.

?? And a machine learning tool called PheNet developed at UCLA Health can quickly identify rare immune disorders from patient electronic health records, significantly improving diagnosis times and patient outcomes

?? Machine learning enables a cheaper and lower power MRI scanner without sacrificing image quality.? This advancement makes MRI more affordable and accessible, particularly for low- and middle-income countries

?? Cedars-Sinai research demonstrates that a deep learning model, trained on over 100,000 echocardiograms from atrial fibrillation cases, can predict which patients may develop irregular heart rhythms within 90 days, potentially improving AFib detection

And there are also very exciting developments in Cardiology:

?? AI can predict heart attacks a decade before symptoms, according to a Caristo Diagnostics study in The Lancet. The AI tool measures coronary inflammation, identifying high-risk patients missed by current diagnostics

?? CathWorks, an AI-powered procedure non-invasively detects artery blockages to prevent heart attacks, determining stent necessity and reducing patient risk

?? And AI can identify heart failure risk by detecting hard-to-spot anomalies, according to a University of Dundee study . Published in ESC Heart Failure, the research analyzed health records and scans from 578 patients, potentially improving early diagnosis and patient care

?? Researchers at the University of Virginia have developed an AI risk assessment tool that uses machine learning to predict outcomes in heart failure patients, enhancing treatment personalization and improving prognoses by evaluating risks of surgery, rehospitalization, and mortality

?? Parkland Memorial Hospital has developed an AI tool to predict and prevent patient violence, protecting healthcare workers. Using EHR data, the tool provides clinicians with alerts to help de-escalate potentially violent situations effectively.

??2) Treatment & Care

AI is proving effective in enhancing treatments patients receive:

?? A study in JMIR Cardio shows an AI coaching program significantly reduces blood pressure in hypertensive adults. Using wearable trackers and BP monitors, the program offered tailored guidance resulting in substantial BP reductions and high patient engagement over 24 weeks

?? An AI app improves diabetes outcomes and reduces ASCVD risk, according to a study. The app significantly reduced HbA1c levels and increased low-risk ASCVD participants compared to standard care, showing AI's potential in diabetes management

?? Twin Health has launched a digital twin AI platform for weight loss, integrating personalized AI and clinical care to help users achieve sustainable weight loss without medication. It promises cost savings by reducing reliance on expensive treatments like GLP-1s

?? University of Florida researchers have developed a machine learning system to identify patients at high risk of discontinuing buprenorphine treatment for opioid use disorder. The AI tool analyzes factors like age, gender, and early treatment adherence to help clinicians intervene early

AI is improving both clinical efficiency and accuracy:

?? Experts at Barts Health NHS Trust used AI to improve clinical efficiency by reading 1,500 outpatient letters, identifying 16 high-risk cases needing action. This AI application aims to ease admin workloads, reduce patient wait times, and improve safety.

?? Cera's AI home healthcare model, could save the NHS ￡1 billion annually by reducing elderly hospital admissions by 50-70%. It predicts falls, tracks symptoms, thus preventing? emergency admissions, and currently saves ￡1 million daily

?? A study shows that an AI model doubles the accuracy of medical professionals reading EEG charts of ICU patients. This interpretable AI helps detect subtle seizure-like events by highlighting relevant brainwave patterns, making it easier for clinicians to identify and understand anomalies.

It is also helping elevate the support patients receive:

?? Researchers at UBC and BC Cancer have developed a new AI model that can accurately predict if a person receiving cancer care will require mental health services during their treatment journey.

?? Artificial intelligence has the potential to help older adults with cancer communicate with their doctors and participate more fully in decisions about their treatment by providing more context or explanation, according to new Cornell research

??3) Latest Approvals

?? FDA adds 128 new AI-enabled medical devices for radiology to its approvals list, with companies like Siemens and GE included. The number of AI/ML devices approved has grown significantly. However, insurance coverage for these devices remains limited

??4) Generative AI

?? Transcarent has launched WayFinding, a generative AI platform designed to help members navigate benefits, clinical guidance, and care delivery. The tool offers personalized insights and easy access to human support, aiming to simplify understanding of health coverage and reduce healthcare costs

?? Darlington College uses AI patients to train healthcare students. These robotic mannequins respond realistically to healthcare scenarios, aiding in patient handling and procedure practice. They aim to enhance critical thinking skills among students.

?? Google's Med-Gemini AI outperforms GPT-4 in healthcare tasks, offering superior clinical reasoning. It excels in medical benchmarks and tasks like EHR retrieval. While showing promise in diagnosing conditions and suggesting treatments, further refinement is needed before real-world use.

?? AI boosts clinical denials appeals success by improving claims management and speeding up the appeal process. It helps prevent denials upfront and generates summaries of patient records for efficient appeals that clinicians can fine-tune

??5) AI Assistants

?? The UK NHS is piloting Asa AI, an AI-driven WhatsApp platform in London to boost cervical cancer screening appointments. The system lets patients book and manage appointments via WhatsApp, using AI to increase attendance and reduce no-shows, easing the burden on primary care teams.

?? A study in JAMA Oncology found AI chatbots provide more empathetic and readable responses to cancer-related questions than doctors. Researchers compared responses from chatbots and six doctors to 200 patient queries, finding chatbots scored higher in empathy, quality, and readability

?? AI chatbots enable surgeons to issue text commands to surgical robots, instructing them to perform tasks like suturing and dilating blood vessels. This technology offers a simpler way to control robots during operations

?? Chatbots are less accurate in non-English health queries , a study by Georgia Tech researchers found. Compared to English, responses in Spanish, Chinese, and Hindi showed an 18% decrease in correctness, 29% in consistency, and 13% in verifiability. The study highlights the need for improving large language models in multilingual accuracy and reliability.

??6) Governance & Ethics

?? UK MHRA sets out a white paper emphasizing the need for UK government regulation of AI in healthcare to ensure safety and accessibility for all. It urges international alignment to avoid losing out on technological advancements. The report highlights the importance of cybersecurity in AI medical devices and calls for transparency, explainability, and fairness in their development and distribution