🚨 Pulse of Innovation 🚨

Breaking news in the healthcare AI

Open AI and Color Health Take on the challenge of implementing AI in Oncology

In collaboration with OpenAI, Color Health has developed a copilot application using GPT-4o to enhance cancer care by interpreting patient data, analyzing healthcare guidelines, and integrating with EHRs to create personalized treatment plans. 

• The copilot application aims to reduce cancer treatment delays, which can increase mortality risk by 6-13% for every four-week delay. It helps clinicians identify missing diagnostics and generate necessary documentation, expediting treatment.

• Clinicians review and refine the AI-generated outputs to ensure accuracy and safety, maintaining high standards of patient care and privacy with HIPAA-compliant data protection.

• Color Health is partnering with UCSF Helen Diller Family Comprehensive Cancer Center to evaluate and potentially integrate the copilot into clinical workflows, aiming to provide AI-generated care plans for over 200,000 patients by the end of 2024

As we have advocated, any AI in healthcare needs to have a “Human in the loop” component. The article says that without AI, the data is fragmented and leads to weeks of delay. Using the copilot takes an average of 5 minutes, and the copilot can identify 4x more missing labs or biopsy results. Further data to corroborate these findings is needed to see this exciting technology's real potential.

🧑🏼‍🔬 Bench to Bedside👨🏽‍🔬

Developments in healthcare AI research and innovations

EHR Shot: An EHR Benchmark for Few-Shot Evaluation of Foundation Models.

 The progress of machine learning (ML) in healthcare has been hindered by the lack of shared assets, unlike the general ML community which has benefited from public datasets, tasks, and models. Addressing these challenges, a new dataset named EHRSHOT has been published, containing de-identified structured data from the electronic health records of 6,739 patients from Stanford Medicine.

Key features:

• Structured data within the deidentified longitudinal EHRs of patients from Stanford Hospital.

• 6,739 patients, 41.6 million clinical events, 921,499 visits, 15 prediction tasks

• The model is available for free: https://huggingface.co/StanfordShahLab/clmbr-t-base 

👩‍⚕️AI in the Clinic 🏥

Real-world and practical use of healthcare AI in clinic

Nuclei.io : Letting Pathologists have full control of AI

• Nuclei.io is a digital pathology AI framework utilizing machine learning and real-time pathologist feedback. It assists in diagnosing digital pathology slides by providing real-time analysis at both the single nucleus and tissue level. Its main advantage is the ability to customize and train to identify images by working with each pathologist's independent data, making it a collaboration between pathologists and AI

 Key Features of Nuclei.io:

1. Real time diagnosis: Rapid analysis of digital pathology slides.

2. Multi-level Analysis: Provides insights at individual nucleus and tissue levels.

3. Improved workflow, efficiency, and quality enhancement: The studies found that pathologists doubled their sensitivity in detecting plasma cells with assistance, improved accuracy, and saved 20% of the time finding lymph node metastases.

   Nature BME Article

🩺 Start-Up Stethoscope 💵 

Trending startups and technologies impacting clinical practice

Dozee: Revolutionising Patient Monitoring with AI-Enabled Beds

In the rapidly evolving landscape of Healthcare Technology, India-based Dozee stands out as an innovative startup that integrates AI to enhance patient monitoring.

The Innovation behind Dozee is an AI-powered contactless health monitoring system embedded within beds aimed at both home care settings and hospital environments. This US-FDA 510 K cleared system leverages cutting-edge sensors and ML algorithms to track vital parameters.

Key Features

• Contactless Ballistocardiography: Utilising mechanical vibrations from body movements, ballistocardiography allows the study of cardiac, pulmonary, and other physiological functions. Dozee's sensors, placed under the mattress, capture micro-vibrations from heartbeats, respiration, skin temperature variations, SpO2 levels, and movement without contact. Contactless BP Monitoring was integrated into the system in 2023.

• AI-Based Early Warning System: This system analyses patient vital trends and automates risk scoring to enhance ward triage effectiveness. The Smart Alerts system provides timely and customizable alerts for each patient, ensuring prompt intervention based on a scoring system.

• Seamless Integration: Dozee seamlessly integrates with existing healthcare systems, ensuring secure data processing and enhancing the monitoring system's utility.

Real World Applications:

•  Home Care: Beyond vital signs monitoring, Dozee beds prove invaluable in understanding sleep physiology and disorders, promising patient-friendly and simplified healthcare solutions for the future.

•  Benefits: Implementing Dozee beds results in enhanced patient care, improved compliance, increased operational efficiency for healthcare workers, and reduced burdens cost-effectively. A recent study conducted at Ramaiah Memorial Hospital, Bangalore, revealed significant outcomes:

  • Increased Time for Patient Care: Nurses reported a 43.11% increase in available time for direct patient care.

  • Enhanced Safety: Healthcare professionals experienced an 89% improvement in safety.

  • Improved Quality of Care: Over 60% of patients noted an improvement in the quality of care received.

• Limitations: Accuracy varies depending upon patient movement and positioning, which is particularly critical in ICU settings, where precise data is crucial for accurate medical decisions.

🤖Patients First, AI Second🤖

Ethical and regulatory landscape of healthcare AI

AI-powered compact MRI could be a game changer for resource-deprived rural hospitals.

• MRI machines require higher electricity consumption to power the magnets that generate 1.5 tesla or higher strength for high-resolution images. Because of these requirements, the use of MRI has been limited in resource-deprived countries such as West Africa, where there are only 87 MRI scanners for a population of 370 million compared to 40 MRI scanners per million population in the USA.

• Now, the University of Hong Kong has developed a newer machine that requires less power and can provide full-body MRI scans with a compact two-unit design. This machine uses two permanent magnets (one above and one below the body) instead of electrified ones. In addition, it uses specialized sensors to limit radio wave noise from the patient’s body and surrounding areas. It has a deep learning algorithm that limits radio wave exposure and enhances the MRI image.

• A recent report on 30 healthy volunteers using this ultra-low-frequency (ULF) machine showed that the quality of the images generated was comparable to that of the traditional machine.

• The ULF MRI machine costs around $22,000 compared to $225,000 for entry-level traditional MRI machines. This low-cost option could be a game changer for resource deprived environments such as West Africa and other such rural areas.