Computer Vision as a Catalyst for Modern Healthcare
By Maria Weinberger, Technology Journalist
For decades, healthcare has been a job for the chosen few who have had the dedication to study hard to gain the top skills which allowed them to make life and death judgments.
On the brink of the Fourth Industrial Revolution, this is about to change, and machines will enhance the precision of specialists. It’s not a matter of replacing humans, but helping them perform better, more quickly, more reliably, and preserving previous results. Right now, when a doctor retires, all the knowledge they’ve gained over time is lost, and the new generation of doctors have to start from scratch.
Big Data, Big Opportunities, Big Problems
The amount of medical data available right now is overwhelming, and this situation will only become worse over the next few years due to the rise of the Internet of Things. Smartphones already monitor a great deal of health and fitness data and wearables are following closely. More sophisticated sensors connected to medical devices will generate gigabytes of data in addition to existing medical imagery.
Such a wealth of data could be good useful if it would be utilized to train algorithms for early disease detection. However, it is not so simple. The most significant challenges are related to privacy, personal data security, and standardization.
Machine Learning and Computer Vision
Machine learning can help with problems that require classification. In the medical field, this is the base of giving a diagnosis. It’s about assessing specific external manifestations and correlations and making deductions about the cause.
It’s about transforming data and probabilistically distributing cases into bins, like “healthy,” “low-risk,” “moderate-risk,” and “high-risk.”
Computer vision focuses on those aspects that require a trained eye, such as classifying skin conditions or accuracy during surgery. While these are jobs that can also be performed by humans, the real value of computer vision for healthcare comes into play when the machine is used to evaluate visual aspects that escape the naked eye such as scanning blood on surgical sponges to assess if a transfusion is necessary.
Applications of Computer Vision in Healthcare
The applications of computer vision for medical purposes are limited only by the human imagination and current development of tech. Applications already in use include tumor classification, CAD systems for surgery, predictive analytics, and therapy.
Medical image analysis
Medical imaging has been around for more than a century in the form of X-rays, dental films and more recently CT scans. The systematic review of the existing images is still developing. Until now it was all part of the doctor’s skill to evaluate the scans or other graphical representations and propose a diagnosis. After this step, the image is never used again in most cases. What if all that information would be stored and the patient’s trajectory followed closely. What if pictures from one medical center could be automatically compared by an algorithm with thousands of similar images from around the world?
This wealth of data needs to be processed by trained neural networks which can classify the information both from a quantitative and qualitative perspective. This is sophisticated research, including computer science, statistics, engineering, biotech, and more.
Predictive analysis and treatment
Predictive analytics is all about minimizing risks, optimizing processes and increasing positive outcomes. In the healthcare industry, this means detecting the possibility of illness before it happens and acting on it. Computer vision can be used together with predictive analytics methods. These include regression analysis, statistical analysis, decision trees, and neural networks. Practically, the visual data collected is converted into numbers, and those are the base for the models above.
This can find applications in medical insurance practices, early cancer detection or deciding between two similar conditions.
Mining and analysis of medical images
Deep learning solutions applied to medical imaging are adaptations of existing deep learning algorithms from other activity sectors. This adaptation requires cross-disciplinary knowledge and is by no means an easy task.
The approach has to be modular, divided between different tasks, such as segmentation of the input information, performing analysis (by regression) and generating new images, as the algorithm learns.
An excellent example of this type of project is the 4D Cardiac. An extensive database of existing images of heart disease patients was used with the aim of training the computer to select images most similar to new pictures from newer patients.
CAD for medical use
Computer vision consultants from Indata Labs state that the technology can be used to create 3D models for educational, training and aiding purposes. VR offers medical students the opportunity to go deep into the human body, down to the atomic level. You can travel through the bloodstream as a red cell or watch neurons create new synapses. All these are possible with a head-mounted set and enough tech to support it and will be far more efficient than pictures in a textbook.
Another use is creating 3D components for dental work and other purposes after a simple scan instead of using age-old technology which required complex additional work, like clay models. This will not only shorten the treatment time, but it can make it more affordable as 3D printers become household items.
CAD for surgery also has a future as a complementary tool for surgeons, acting as a helper or warning when the operation performed is too dangerous for the patient. The long-term goal is to have robo-surgeons or to perform telesurgery through robotic arms with top precision.
Assisted Living, Rehabilitation, and Training
Computer vision will also upgrade the care market. It will provide people with physical disabilities the opportunity to enjoy life in a more accomplished way. Smart wheelchairs, better head-mounted sets for the blind or even a personal assistant for the elderly can all be powered by computer vision.
The same systems are also useful for patients recovering from accidents which can be guided by a smart method to use the right posture. Athletes in training can have a robo-trainer which monitors not only their overall performance but also looks at small differences such as the angle of their muscles or the strength of their throw.