New collaboration with Olympus offers a unique opportunity to collect data in 3D from endoscopies of the intestine. It is a leaping point for developing a feedback system that automatically and in real time is capable of assessing the validity of the examination and thus increasing patient safety.
The two PhD students, engineer David Norsk from DTU Compute, who is specialising in mathematical modelling, and medical doctor Andreas Slot Vilmann from the Capital Region and the University of Copenhagen, are working on the same research project, but with two different approaches. They are developing an automatic feedback system that can assess how skilled a doctor or nurse is to perform an endoscopy of the intestine (colonoscopy). The system must ensure that the examination is performed correctly and thus results in less discomfort and greater safety for the patient.
From 2D to 3D Data
During autumn 2017, the project took another step toward realising that vision. Here, they entered into an agreement with Olympus, the company behind the equipment used by many hospitals when performing colonoscopies. The agreement means that Olympus now makes six prototypes available to the two researchers in order that they may capture raw 3D data from the colonoscopies.
“Earlier, we were able to see the tip of the coloscope in 2D, but now that Olympus has joined us, we can receive the raw X, Y, Z coordinators five times per second from points evenly distributed on the entire coloscope. This means that we can constantly see the whole movement of the coloscope in 3D”, David Norsk recounts enthusiastically.
3D Data from Simulated Colonoscopies
In December and January, Andreas Vilmann obtained such 3D data from 12 experienced colonoscopy doctors and 12 novices who have performed numerous colonoscopies on simulation dolls. Now it is time to find contexts and patterns in these data.
“In the basement under DTU Compute we are building a system that – using some smart mathematics – will be able to establish the connection and find out how to separate novices from experts. We use machine learning to see what describes experts and what describes novices”, explains David.
In other words, when they have enough data to characterise a good examination, they will have some concrete measurement parameters that may be used to judge the quality of all colonoscopies.
The next step is to retrieve data from proper studies at the clinic. Therefore, they have built an automatic recording device that may be connected to the colonoscopy equipment. It automatically collects data about the examination as soon as it is switched on. This means that in spring, Andreas Vilmann will be able to automatically and continuously collect data from six hospital wards without having to be present.
Guide to Preventing Painful Loops
However, it is not sufficient to study how good doctors perform the examination. 10-20% of all colonoscopies are difficult to perform for instance because loops are formed on the coloscope which stretches in the intestine and hurts the patient, at the same time halting the progress of the examination.
“You actually do not know to what extent loops are the culprit. We are going to measure whether it is indeed loops, what kind and where they are so we know what to do to prevent them”, says David Norsk.
He uses mathematical models to describe the folding of proteins to identify and characterise loops during a colonoscopy.
“The math describes how much the coloscope twists around itself, simply to give us a measurement for how twisted the coloscope is”, he says.
Under simulation training, this new insight may be used to guide doctors who are pursuing this education to loosen the loop and proceed with the examination. In the clinic it may provide information on the extent to which loops occur and what the consequences are for the examination time. When the system is fully developed, it will be able to predict whether loops are about to be formed in real time. It will help the doctor prevent the forming of the loop, thus ensuring that the patient receives a more effective treatment with fewer complications.
Quality may be Measured as Steady Progress rather than Time
The actual examination time is currently the best measurement parameter for the quality of an examination. When the coloscope is led all the way through the rectum and the large intestine to where the small intestine begins, the examination of the intestine starts, while slowly retracting the coloscope. This retraction is estimated to take 6.5 minutes. However, the two PhD students do not believe that time is a satisfactory target for the quality of the examination.
“We can set some more analytically based quality goals. Mathematically, it is possible to distinguish between entry and retraction in the data sets, and we can thus measure the quality of the examination itself to see whether there is a continuous movement throughout the entire intestine. In principle, we will also in the long run be able to look at data to see if you still have to investigate small areas of the intestine”, says David Norsk.
“The prototype from Olympus is the salient point which makes this specific thing possible”, he emphasises.
Lars Konge, clinical professor from the Copenhagen Academy for Medical Education and Simulation (CAMES) at Rigshospitalet, is a supervisor of the project, and he is excited about the perspective of being able to assess the quality of the examination.
“We are moving toward the point where we can develop a virtual colonoscopy supervisor – a monitoring system that ensures ongoing quality. It was a distant dream when we started. But it’s closer to realisation now. With the new results, we will be able to ensure that the examination may progress without interruption. If there is no progress, a more experienced doctor may come and take over the procedure. We can also hold an exam, on the simulator and in real time, insuring everyone reaches a satisfactory level – this will increase the patient security!” emphasises Lars Konge.
The new monitoring system will have a high growth potential as it can be used by hospitals worldwide.