How and when can visualisation of patient data offer real added value to diagnostics, and even become indispensable? Dr. Lelieveldt believes that the role of visualisation techniques applied for daily clinical routine is growing in interest within the medical community. More and more imaging modalities are being used to generate enormous volumes of data, like for instance in cardiac MR imaging. One patient examination delivers 200 to 400 images but current trends in research require 2000 to 5000 images during one scan session in order to try and map the patient's complete heart function. To present this huge amount of data in an orderly manner for diagnostic purposes, visualisation techniques are absolutely necessary.
"Knowledge through measurement" constitutes another major axiom in medical science, as Dr. Lelieveldt explained. For example, the visualisation of an exact quantitative analysis of obstructed blood vessels can help the surgeon to select the right procedure for intervention. In fact, there are four major application areas in which medical visualisation techniques are of immediate and practical use at the LUMC, including segmentation and analysis; motion pattern analysis; simulation of shoulder movement in collaboration with the Technical University of Delft; and interactive anatomical atlas models for the education of medical students.
At the LUMC Image Processing Laboratory, slices are generated from MRI cardiac data during the different time steps of one heart beat in order to facilitate quantitative analysis. This research allows to visualise the pump function of the entire muscle or to view and measure fragments of the heart wall. To this purpose, you need organ surface rendering to calculate all relevant data for the diagnosis, as Dr. Lelieveldt showed. Manual segmentation however is too time consuming and therefore, several techniques have to be developed to detect the organ surface automatically. Predefined anatomical knowledge is required to design this type of software since there are important morphological variations when comparing one patient's heart to that of another patient.
Dr. Lelieveldt's team in Leiden is using "active appearance" models for segmentation. From a set of examples, it is possible to generate one active appearance of an "average heart", as well as to calculate individual variations to visualise the differences from patient to patient. When a new data set has to be analysed, the "average heart" can be used to fit the heart segment of the individual patient and comparative analysis between the two results in a diagnostic solution which is statistically plausible. In this way, it is possible to automatically calculate the different time steps of the entire heart beat. Correct visualisation is absolutely impossible without preliminary segmentation, as Dr. Lelieveldt stressed.
Motion pattern analysis forms an ideal topic for visualisation. At the LUMC, a 3D study out of MRI cross-sectional data has been made of the wrist movement in which the different bones of the joint can be viewed in full action. Unstable movement behaviour can be compared with normal motion and sub-movements can be analysed. It is even possible to fix one single bone and view the effect of this intervention on the other ones, as Dr. Lelieveldt demonstrated. Simulation has been experimented with in the Dutch shoulder group, which applies visualisation techniques to simulate shoulder movements in a biomechanical model in order to calculate and predict the involved muscle forces.
In this application, visualisation is utilised to interpret the simulated data. Patients with an implanted shoulder prosthesis often need to undergo surgery to move atrophied muscles to another spot in the joint. The simulation visualises which motion still remains possible if one of the muscles for instance is being transferred to the shoulderblade. In conclusion, Dr. Lelieveldt showed some examples of interactive 3D anatomical atlases for educational purposes. These models, based on the Visual Human Project, are also applied in the radiological practice to demonstrate what organs look like when specific imaging modalities are being used. With a simple mouse click on the 3D model, you can even obtain related background information.
In all four application areas mentioned by Dr. Lelieveldt, visualisation offers practical and relevant support to clinicians and medical students. However, it is difficult to convince more conservative physicians to utilise visualisation in daily clinical practice. Meanwhile, the number of visualisation applications is still growing. Once new visualisation techniques have left the development phase, they should be introduced to the physicians, according to Dr. Lelieveldt. After all, it are the physicians who will promote these innovative applications and above all are going to use them.