Four ways computer vision improves patient care
Olga Ezzheva, a technology writer for software engineering and IT consulting services Oxagile, explains how computer vision can help medical professionals improve patient outcomes.
Computer vision is a hot subfield of artificial intelligence, which aims to give machines a human-like capability — to see and understand the world around us. And with rapid advancements in deep learning and growing computing power, computers cope with the task at hand more and more confidently.
Through acquiring, processing, and analysing, computer vision (CV) enables machines to capture visual images, gain an understanding, and deliver a meaningful result. Highly accurate real-time object recognition benefits a multitude of industries — from self-driving cars to industrial robots to security applications.
Rich in clinical imaging data, healthcare too harnesses the power of computer vision to support decision making for medical professionals and improve patient outcomes. Here, we will walk you through four ways computer vision supports healthcare.
Automated monitoring for better hand hygiene
Hospital-acquired infections (HAI) are a serious safety issue. World Health Organisation reports that out of 100 hospitalised patients, seven in developed countries will acquire an HAI. For developing countries, the numbers are even higher. These infections, also known as nosocomial, cause higher morbidity and mortality, leading to significant healthcare costs. In the United States, the economic burden estimates range from $28 billion to 45 billion.
Hand hygiene is the first and most important measure to control the spread of infections. And hospitals are putting a lot of effort into implementing strict hand washing protocols throughout their units, but monitoring hand hygiene compliance is challenging without proper tools.
Stanford Partnership in AI-Assisted Care (PAC) took on the challenge and developed a smart CV-based solution that automatically tracks medical staff around the hospital. The system leverages deep learning to identify hand hygiene activity and detect any missed hygiene events for real-time intervention.
CV-based approach to more accurate diagnoses
X-ray, CT scans, and MRI images yield valuable information that a clinician needs to diagnose a disease. While manual analysis of this clinical imaging is time-consuming and prone to errors, smart computer vision-based systems interpret medical images much faster and mark suspicious sections to alert a doctor.
Trained with deep learning algorithms, CV models can also automatically detect diseases. A system built by the scientists from Stanford University uses cutting-edge deep learning algorithms to spot Alzheimer’s disease biomarkers from MRI data with 94% accuracy, outperforming other methods.
Another promising use case for a CV-based approach to better diagnosis is cancer detection. By training a convolutional neural network (CNN) with over 100,000 images, researchers were able to reach 95% accuracy in skin cancer detection — better than a panel of medical experts.
Fall detection systems for the elderly
Each year, 37.3 million falls occurs that are severe enough to require medical help. Though most falls are non-fatal, they still have an adverse impact on public health causing head injuries and broken bones. With age being a key risk factor, falls become a major cause of morbidity and mortality among adults aged 65 and older.
Although the elderly are at the highest risk of suffering fall-related injuries, less than half are willing to discuss their falling with care providers. Moreover, if a fall happens at home, it can be hours before someone finds a senior person — hours that could be critical.
To ensure timely provision of medical help, researchers leverage computer vision to build a fall detection system that monitors daily activities and automatically alerts a caregiver. The CV-powered system relies on advanced deep learning techniques for image segmentation, frame-to-frame object tracking, and state classification, which allows the solution to identify a fall with 96% accuracy.
Image-guided surgeries
Computer vision has also found its way into the operating rooms and is now helping surgeons improve their practice through better surgical planning and greater operation accuracy.
In image-guided surgery, also known as computer-assisted surgery, CV algorithms process preoperative medical images and convert them into a detailed 3D model of a patient’s anatomy. The subsequent image-to-patient registration allows surgeons to see the hidden structures and track their medical instruments for safer yet more precise navigation during an operation.
One of the health tech companies that advance computer-assisted surgery technologies is 7D Surgical, a Toronto-based medical company. Its Machine-vision Image Guided Surgery (MvIGS) system is designed for fast and secure spinal navigation through real-time visualisation of medical tools. With patient registration under 20 seconds, the guided surgery becomes faster, while also being radiation-free.
Wrapping up
Computer vision is already driving better patient care through more accurate diagnosis and early detection of diseases, 3D visualisation and precise surgical guidance, and even improved monitoring of daily activities for fall detection. Impressive as they are, these applications of CV in healthcare are just scratching the surface of the tech’s potential. And as the technology continues to mature along with advancements in AI and deep learning, more practical use cases will be coming our way.