KASI - Knowledge Assited Sparse Interaction for Peripheral CT

Diagnostic imaging has evolved from simple capture and visualization of individual X-ray projections to the complex ability to capture and visualize the true three or even more dimensional inner structure of the diagnosed body. This not only made possible to see the unseen earlier, but also significantly increased the amount of data to be processed. Now, often thousands of images have to be inspected in a single diagnostic session. This trend inevitably leads to the development of new data processing and data visualization techniques which have materialized in the tools used on diagnostic workstations. In order to cover all ever growing application areas of medical imaging, such tools offer numerous options for both preprocessing and visualization. Thus, the simplicity of handling the old X-ray images is getting lost in favor of generality - but unfortunately also giving in to complexity, which makes the new possibilities less accessible due to the increased requirements in training. Our research, performed in the framework of the two FWF funded projects (FWF P-15217, April 2002 - May 2004 and L291-N04, April 2006 - July 2009), showed that the trend towards growing complexity is not the only option. We developed a unique processing pipeline, in which data processing was separated from diagnostic reading. Here, processing was performed by trained technologists. Thus, the radiologist`s task was reduced to reading of images rendered off-line according to specified and optimized requirements. Our thorough clinical tests have shown that this approach in the diagnostics of the peripheral artery occlusive disease (PAOD) by CT angiography delineates the disease accurately and contributes to its treatment planning. Our approach not only proved to be efficient, but was also well accepted by the radiologists due to its simplicity and intuitiveness. However, in certain situations it becomes a limitation since no interaction is possible. Therefore, in our future work, we want to develop a new data centered approach to the interaction with visualization tools, which would open the ever growing plethora of visualization techniques and data types yet in an intuitive and predictable way to the user. We call this new interaction paradigm knowledge assisted sparse interaction (KASI). Here, we want to put the data rather than the tools into the focal point. A list of available analysis and visualization tools will be constructed on the basis of knowledge about the problem at hand and will be offered to the radiologist in a context-specific way. Thus, the radiologist will be offered only a sparse set of important interaction controls, and will have the possibility to concentrate on his/her primary duty-data evaluation and radiological reading. Except for the development of the basic KASI interaction framework and its implementation and clinical evaluation in the diagnostics of the PAOD by CT angiography, we want to extent the existing processing and visualization techniques and tools towards exploitation of new emerging diagnostic technologies, mainly of dual- valued CT data provided by the new dual energy and dual source scanners. We expect that this technology augmented by sophisticated statistical reconstruction techniques will improve the diagnostics of thin vessels with calcifications and stents which is currently a problematic area. To accomplish these goals, we propose an interdisciplinary approach with one clinical site, two technical development sites in the field of computer graphics and medical imaging, and one international collaborating site at Stanford University (USA). Such interdisciplinarity will provide a fast clinically relevant feedback, which would allow us to incorporate the necessary features and changes already early in the research and development stages. We expect to complete the research and implementation within a three year period.

Peripheral arterial disease is a manifestation of atherosclerosis, which is an increasing health problem in the western world. Revascularization by means of endovascular therapy such as stenting, balloon dilatation, or bypass surgery has improved the patients outcome over the last decades. However, for an accurate treatment planning, a complete depiction of the peripheral vessels including the in- and outflow is necessary. Computed tomography angiography (CTA) was shown to fulfill this task with a high spatial resolution, but leads to thousands of axial images, which are often hard to interpret even for skilled radiologists. Thus, computer supported analysis of blood vessels for their diagnosis and treatment is an important research field in radiology and, in order to optimally decide on the therapeutic procedure, specific computer supported diagnostic methods are required. They assess the health state of blood vessels and answer clinically relevant questions such as, e.g., if blood is partially or entirely hindered from flowing through a vessel by a clot or a calcification.Throughout this research project, several techniques have been developed for assessing the interior (or lumen) of blood vessels by virtually cutting along their centerline. This allows radiologists to precisely judge if blood is able to flow through a vessel or if it is significantly hindered, as in the case of a stenosis. With the visualization technique called Curved Surface Reformation (CSR), high-quality interactive visualizations of the vessels' interior can be achieved by computing the vessel lumen entirely in three dimensions. This enables the assessment even of complex blood vessels in the brain. Other innovations dealt with the reduction of the vast numbers of images to inspect by presenting only the most relevant visualization techniques to physicians or by aggregating the images around the centerlines of the vessels themselves. This provides a single image as overview and guidance map.This research was carried out in an interdisciplinary environment involving both radiologists and computer scientists. Within this environment, the clinical usability of the new reformation techniques was confirmed which enabled us to integrate the new tools into the infrastructure in the daily routine of a radiological department of a hospital. Altogether 600 cases are currently processed by these tools per year at various hospitals. However, the diagnostic accuracy of these reformation techniques was not fully evaluated yet, but is currently being investigated in several ongoing clinical trials. Encouraged by our preliminary results, we intend to evaluate our new workflow not only for peripheral arteries, but also for carotid and coronary arteries, being the source for cerebrovascular and coronary artery disease, which are additional manifestations of atherosclerosis.