🚨 Pulse of Innovation 🚨

Breaking news in the healthcare AI

AI solutions improving accessibility, mobility and assistance 

The recently concluded “AI for Good” summit, organized by the International Telecommunications Union in partnership with the UN has “Good Health and Well Being” track, a sustainable development goal, that showcased several AI solutions impacting our lives including improving patient mobility, accessibility, advancing health monitoring and providing assistance for rehabilitating patients.

• Autonomyo is an AI powered robotic rehabilitation platform that supports exoskeleton of neurologically debilitated patients and supports their recovery including gate improvement.

• Bionik focuses on bionic AI powered customized limbs that do not require surgical implant providing more cost-effect solution to millions.

• INBRAIN’s brain computer interface (BCI) provides machine learning based platform that not only helps better understanding of neurodegenerative diseases but also helps patients suffering from such diseases the Parkinson’s with more personalized treatment.

• Lio Robot is a multipurpose robot that is primarily designed to provide assistance in healthcare setup. It can perform variety of activities from fetching a water bottle for patient to interacting the with patients in variety of ways and uses AI to anticipate the tasks in different settings. In addition the robot is very useful in assisted living facilities as well as in home setup for the aging disabled population.

• 

🧑🏼‍🔬 Bench to Bedside👨🏽‍🔬

Developments in healthcare AI research and innovations

Enhancing EEG Pattern Classification in ICUs with Interpretable Machine Learning: The ProtoPMed-EEG Study

Study Overview

• Electroencephalography (EEG) is a crucial tool for monitoring brain activity, particularly in patients with epilepsy or other neurological conditions. Accurate classification of EEG patterns into ictal (seizure activity), inter-ictal (between seizures), and injury-related patterns is essential for diagnosis and treatment planning. However, interpreting EEGs is challenging and requires significant expertise. Machine learning (ML) offers a potential solution by enhancing clinical performance through automated, interpretable models.

• A recent study “Improving Clinician Performance in Classifying EEG Patterns on the Ictal-Interictal Injury Continuum Using Interpretable machine Learning” evaluates clinician performance in terms of accuracy, confidence, and speed of EEG interpretation with the help of ProtoPMed-EEG which is a prototype for precision medicine in EEG analysis.

  Key Findings

• Model Performance: High accuracy in distinguishing between ictal, inter-ictal, and injury-related EEG patterns.

• Impact on Clinicians: Using this tool significantly improved performance among users from 47% to 71%. The average time taken to make a decision was longer with AI assistance than without (32±33 seconds vs. 25±39 seconds). An increase in mean inter-rater reliability (IRR) was observed.

• Summary: The study demonstrates that interpretable ML models can significantly aid clinicians in accurately classifying EEG patterns by enhancing both the performance and the trustworthiness of automated systems, such models can become valuable tools in clinical settings.

👩‍⚕️AI in the Clinic 🏥

Real-world and practical use of healthcare AI in clinic

Seismometer: EPIC EHR’s open source tool to democratize the AI application validation

The EPIC, one of the commonly used EHR systems has provided an open source software with EPIC integration that can evaluate the machine learning and AI models implemented by hospitals and provide its real-time performance information. HAIP network, Duke and University of Wisconsin are important partners in this EPIC initiative.

Functionality

Smaller hospitals and health systems may not have tools to accurately validate any machine learning or AI platforms using their own local data. The EPIC Seismometer can use the local data from the hospitals and provide information on performance on their own machine learning and AI models’ performance. The software us open source and is available on GitHub. It is an excellent resource for small centers as it uses their local data a huge validation parameter.

Transparency

The Seismometer has a Fairness audit tool which focuses on providing standard set of analysis that is much more broadly accessible to different centers. This will provide all the necessary parameters to understand the performance of the models albeit in much more standardize way very similar to getting lab test done at different labs. This increase widespread transparency.

Regulatory Compliance

Several rules have been enacted are to be implemented within next 1-year on use of health AI including the rule that ensures no discrimination has been done with use AI models. These EPIC Seismometer will identify such models and help hospital improve their compliance to meet the industry standards.

🩺 Start-Up Stethoscope 💵 

Trending startups and technologies impacting clinical practice

VitVio : Enhancing Operating Room efficacy with AI

Operating rooms are high-revenue environments where optimizing patient safety and precision while minimizing costs and wastage is crucial. VitVio, an AI company, aims to improve administration and increase efficiency, safety, and profitability in operating rooms.

How It Works: VitVio uses AI and computer vision technology to analyze video feeds from cameras in operating rooms. The system functions as follows:

• Video Capture: High-definition cameras record the entire surgical workflow, including staff movements and equipment usage.

• Computer Vision Analysis: AI algorithms analyze the video data in real-time to track objects, people, and actions, including surgical tools and procedural steps.

• Data Processing: The data generates insights on surgery progress, tool usage, protocol adherence, and operational efficiency metrics.

• Workflow Optimization: Insights provide real-time estimates on surgery duration, identify bottlenecks or safety risks, and suggest staff and equipment optimizations.

• Training and Guidance: The system uses recorded surgical videos to offer tailored training based on actual procedures performed at the hospital.

  Positives:

• User-friendly integration and Real-time reporting

• HIPAA and GDPR compliant, with end-to-end encryption and data anonymization

  Concerns:

• Privacy and security issues with intra-op video footage

• Need for multi-center implementation to enhance algorithm accuracy and address scalability and affordability

🤖Patients First, AI Second🤖

Ethical and regulatory landscape of healthcare AI

Patient Safety in Health AI: Focus on Transparency

• The world of AI is moving fast taking a generational leap with use of generative AI. This enabled several audio-visual platforms to provide relevant content especially in the healthcare market from patient engagement and diagnostic tools to administrative and academic applications. All these wonderfully displayed products are not necessarily transparent about the functioning of their product/platform.

• To counter this issue of transparency the government enacted the Health Data, Technology and Interoperability Act (HTI-1 rule) or in other words Health AI transparency act, in December 2023, updated in February 2024 and finalized in March 2024.

• Under the new rule, AI vendors must share information about how their software works and how it was developed. That means disclosing information about who funded their products’ development, which data was used to train the model, measures they used to prevent bias, how they validated the product, and which use cases the tool was designed for.

• The Office of the National Coordinator for Health Information Technology (ONC) director Mickey Tripathi recently gave update on this rule at the Reuters Digital Health summit suggesting that with the new role clinicians and administrators will have more relevant information about the AI product before its use and while it is being used under the EHR system. This will lead widespread adaptation of the AI with more confidence. Several key outlines are given here.