Computer Vision
Our participation in the Stanford Partnership in AI-Assisted Care, which is co-led by CERC Founder and Director Arnold Milstein and world-renowned computer scientist Fei-Fei Li, applies cutting-edge computer vision technologies to efforts to boost the reliability of clinical care. Sensors passively capture data from the clinical environment, while machine-learning algorithms are developed to automatically detect patient and staff activities.
CERC has also demonstrated how ambient AI can detect the failure to mobilize ICU patients as intended. Enhancing mobilization will help prevent permanent cognitive deterioration and debilitating muscle weakness among seniors following ICU care.
Projects and Outcomes
- To support seniors' quality of life, extend their ability to live at home, and help prevent their relocation into institutional care, we are designing an integrated solution to detect and record daily activities to identify triggers of clinical deterioration before patients require hospital care.
- To help reduce the rate of hospital-acquired infections, we are investigating the use of multiple sensors for the detection, measurement, and evaluation of hand hygiene in hospitals. Our sensors are deployed at two major healthcare partners: Intermountain Healthcare and Lucile Packard Children's Hospital (LPCH) at Stanford.
- Activity detection in intensive care units can replace the expensive, time-consuming manual process of monitoring and logging currently performed by nurses and other healthcare staff. Our goal is to design a system which automatically provides an annotated list of all daily ICU activities, leading to quicker and safer patient recovery and reducing costs.
- In collaboration with the Stanford Department of Surgery, we are using computer vision technology to assess surgical technique in the OR. This work will be used to attempt to relate markers of surgical technique with clinical outcomes.
We have also applied an automated visual system to predict both burn severity and spatial outlines. Our researchers have been able to achieve an impressive 85% pixel accuracy in discriminating between burnt skin and the rest of the image, with clear pathways toward even stronger results.
Our AI program has drawn interest from large insurers that seek to improve the efficiency of health care systems serving large groups of their health insurance enrollees.
Publications and News
- – 7/12/21 - Nature
Using deep learning to identify the recurrent laryngeal nerve during thyroidectomy
This CERC PAC research suggests the potential for deep learning to provide insight to the surgeon about image capture technique for achieving high performance.
- – Journal of Medical Internet Research
Parents’ Perspectives on Using Artificial Intelligence to Reduce Technology Interference During Early Childhood: Cross-sectional Online Survey
A survey by CERC revealed that most parents, particularly younger ones, said they’d welcome AI tools to help mitigate technoference during interactions with their children.
- – JAMIA
Automatic detection of hand hygiene using computer vision technology
As part of CERC’s Partnership for AI-Assisted Care, researchers concluded a computer vision algorithm was equal to human observation in detecting adherence to hand hygiene protocol
- – The Lancet
Computer Vision's Potential to Improve Health Care
Milstein and Topol: Eventual adoption of AI-computer vision assistance could help reduce medical error, though it faces many hurdles.
"Early-stage partnership across operations, providers, and data scientists is critical for developing and implementing AI in healthcare, and the work we are conducting with CERC has the potential to improve patient care and advance this important field of research."
Margaret Ann Smith
Director of Operations
Healthcare AI Applied Research Team
Department of Medicine, PCPH
Stanford School of Medicine