At CAMES AI we work with artificial intelligence in medical education. We research and develop AI technology that supports and improves learning processes, decision-making and daily work in hospitals.
What we can help with
Automated competence assessment
We develop new methods for automated competence assessment via AI and simulation. For example, using 3D-printed phantoms or AI-based competency assessment.
Testing clinical AI systems
By using medical simulation, we can help test clinicians' interaction with AI solutions as a pre-test before implementation in the clinic.
Advice and assistance
We can help think education into AI development to ensure delivery of the right type of AI support to the right clinicians - at the right time.
Current AI projects
AI-based training in the diagnosis of birthmark cancer
Through training using a mobile application, the project aims to make it more efficient for doctors to learn how to diagnose skin cancer. What we also call "Intelligent" learning in pattern recognition.
It currently takes more than 6 years to become proficient at telling the difference between benign and malignant skin tumours. This leads to delayed diagnosis of skin cancer and high costs for the healthcare sector due to unnecessary removal of benign tumours. The reason for the slow upskilling of doctors is probably that it is difficult for a young doctor to get visual feedback on a large number of skin tumours of different types.
We're working to improve this by giving doctors and nurses access to training in skin diagnostics via the learning app - Dermloop Learn. (See photos from app store)
The app exposes the doctor to a huge library (20,000+ cases) of images of previous skin tumours and their diagnosis. Doctors are guided through a learning process optimised by artificial intelligence, which continuously measures the doctor's multi-dimensional competence and selects the optimal learning material based on this.
Results and effects:
As part of the research project, 76 medical students with no previous experience were trained in skin cancer diagnosis for 8 days. And they reached the same level as doctors with 3-4 years of experience. Combined with a tele-dermatology extension of the system, there is plenty of potential - in terms of economics, skills development and patient safety. For example, the research group behind the project has estimated that up to €800 million could be saved in healthcare. DKK 800 million a year if the technology were implemented nationally and used by all general practitioners, dermatologists, plastic surgeons and pathologists.
AI for improved diagnosis of pregnant women
In the SONAI (Sonography AI for pregnancy) project, we are working to develop AI models that support clinicians' performance when performing ultrasound scans. In Denmark, all women are offered an ultrasound scan at around 12 and 20 weeks of pregnancy as part of their routine antenatal care. Ultrasound is an important tool for examining pregnant women and their unborn children. We use ultrasound, among other things, to check for malformations in the unborn child around week 20.
Undetected congenital malformations of the heart, for example, can have serious and lifelong consequences. And although we're good at ultrasound scans in Denmark, we don't find all serious malformations and conditions - partly because it takes years of training to get good at these scans.
The aim of the SONAI project is to improve the diagnosis of pregnant women by providing AI-based feedback to clinicians performing ultrasound scans.
The project has been selected by the Danish Regions to be an AI Signature Project and has been awarded funding from the Danish Finance Act as well as a number of Danish foundations, including the Innovation Fund (DIREC), Aasa and Ejnar Danielsens Fond, Johannes Fogs Fond, Dagmar Marshalls Fond, RH-OUH forskningsfonden, and Region Hovedstaden Innovationspulje. The project is the result of a strong collaboration between CAMES and DTU Compute and the Department of Computer Science at the University of Copenhagen.
AI for training and assessment of invasive fetal medical procedures
In this project, we are developing an AI-based instrument for assessing competencies when it comes to performing invasive fetal medicine procedures.
Every year, nearly 60,000 pregnant women attend a first trimester screening in Denmark. Of these, about 6% are found to be at increased risk of a chromosomal abnormality or genetic anomaly in the fetus. They are therefore offered prenatal genetic invasive diagnostics in the form of either amniocentesis (AC) or chorionic villus sampling (CVS). Involuntary pregnancy loss is a serious complication of CVS and AC. In most major studies, this risk is very low, as CVS and AC in skilled hands are extremely safe procedures. However, figures from the latest national audit in Denmark show that the risk varies across the country and can be up to 3%. It is known that operator experience influences the risk of serious complications, but we still lack knowledge about which aspects of the operator's work pose this risk.
The aim of the project is to use AI models to understand what distinguishes excellence in performing CVS and AC. In addition, the aim is to ensure that everyone performing the procedure has the right skills.
(simulation setup for the project)
Acute caesarean section and AI
An emergency caesarean section is performed when the birth does not go as planned. Complications during labour can be life-threatening for mother or baby, and in these situations it is necessary to deliver the baby as soon as possible.
The quality of patient care in emergency caesarean section depends on many factors. For example, the indication for the caesarean section, the degree of danger to mother or baby, the mother's BMI, previous abdominal operations, psychological factors of the parents-to-be, the size of the baby and the type of anaesthesia can all have an impact on the quality of patient care. Furthermore, the individual, collective and relational competencies of the team will influence both professional quality and patient-perceived quality.
Using AI models, it is possible to identify patterns in large data sets where multiple diverse factors need to be investigated simultaneously. Furthermore, AI models are able to explore the relative importance of the factors. This knowledge is particularly important when planning future teaching and quality improvement initiatives.
Hip fracture surgery and artificial intelligence.
A project using image data to develop AI algorithms that can help orthopaedic surgeons choose the most optimal surgical methods for hip fractures and ensure the highest possible quality of surgery.
Artificial intelligence and automated skills assessments
Competence monitoring in the clinic is often resource-intensive - especially if it has to be carried out in a structured and continuous way.
CAMES has been running simulation-based courses in osteosynthesis of hip fractures based on validated competency scores for the last 7 years. The courses have provided several thousand simulated radiographs based on training and testing by both novices and experts.
This project seeks to uncover whether training an AI algorithm using deep neural network can automate competency assessment of performed surgery in the simulated context as well as on clinical per- and postoperative radiographs. Such a tool could monitor the quality of performed surgery at both the individual and organizational level and allow for earlier intervention in case of competence and quality gaps.
We have a close collaboration with DTU Compute and DIKU (Department of Computer Science, University of Copenhagen). In addition, we have a large number of "customers" who use our services, facilities and expertise. Please contact us if you see a collaboration between you and CAMES AI.