An image processing based patient-specific optimal catheter selection

Coronary angiography is performed to investigate coronary diseases of the human heart. For better visualization of the arteries, a catheter is used to inject a contrast dye into the coronary arteries. Due to the anatomical variation of the aorta and the coronary arteries in different humans, one common catheter cannot be used for all patients. The cardiologists test different catheters for a patient and select the best one according to the patient's anatomy. To overcome these problems, we propose a computer-aided catheter selection procedure. The basic idea of this approach is to obtain MR/CT images before starting angiography. From these images, the patients' arteries are segmented and some geometric parameters are computed from the segmented images. At the same time, geometric parameters are computed from the available catheters. A model is developed, which is based on these parameters from the patients' image data and parameters from the catheters. This model reduces the number of catheter choices. In the next step, the reduced number of catheters are simulated and the most optimal catheter is obtained. A series of validation tests were conducted for segmentation, geometric parameters' estimation, parameters based catheter selection and simulation model. In our experiments, we compared catheters selected in the clinic with the catheters suggested by the image processing based model. For these experiments, the ground truth data were obtained from the clinical partner. In the clinic, angiography of twenty four cases was performed. An experienced cardiologist selected catheters based on his experience and knowledge in the field. In the next step, CT/MR image data that was acquired prior to the angiography was used for the image based catheter selection model to find optimal catheters. For every patient, three most optimal catheters were suggested by the model. These three optimal catheters were ranked as first, second and third ranked catheters. Catheters suggested by the model were compared with the catheters selected by the cardiologist. It was found that in 41% cases, model based top ranked suggestions were the same as that were used in the clinic. In 25% cases, the catheters used in the clinic were the model's second ranked catheters. In 21% cases, the catheters used in the clinic were the model's third ranked catheters. In 13% cases the catheters used in the clinic were not in the list of suggested catheters. In further experiments, the clinicians graded catheters based on catheter's performance and placement in the arteries. Most optimally placed catheters were assigned good grades, and less optimal catheters were assigned bad grades. It was seen that the model suggested similar catheters to the clinically good graded catheters but suggested different catheters to the clinically bad grade catheters. All these experiments showed that the method of an image processing based catheter selection is clinically applicable, and the only requirement is to have patient's image data before starting the angiography. It was shown that this tool will be of great help for the experienced as well as the non experienced cardiologists to have a catheter suggestion before starting the angiography.

In dieser Arbeit haben wir Modelle präsentiert, die von unterschiedlichen MR- und CT-Bildern Information sammeln und patientenspezifische Informationen für die Behandlungsplanung und Diagnose bereitstellen. Die Beiträge sind Modelle zum Sammeln nützlicher Information von unterschiedlichen Bildern, Modelle zum Planen von Angiographien und simulationsbasierte Entscheidungsunterstützung vor der eigentlichen Prozedur. Unser erster Beitrag ist das Sammeln von Informationen aus verschiedenen 3D-Bildern. In dieser Arbeit haben wir eine Idee vorgestellt, um Bildinformation durch Segmentierung und Registrierung zu fusionieren. Dies ist das erste Mal, dass ein vollständig automatisches Segmentierungsmodell für die Aorta und die Koronararterien präsentiert wurde. Der zweite Beitrag liegt im Gebiet der quantitativen Messung von verschiedenen klinisch relevanten Parametern. Ein 3D-Modell wird aus den 3D-MR- und -CT-Bildern erstellt, und dann werden die klinisch relevanten Parameter automatisch berechnet. Der wichtigste Beitrag dieser Arbeit ist in dem Gebiet der Katheterangiographie und Katheterauswahl. Unterschiedliche Modelle wurden definiert für die sehr schwierige Aufgabe der Katheterauswahl. In dieser Arbeit wurden verschiedene bildverarbeitungsbasierte Modelle für die optimale Katheterauswahl definiert. Diese Modelle sind in der Lage, die gängige Selektionsstrategie durch Versuch und Irrtum zu ersetzen. Es wurde gezeigt, dass die Methode klinisch anwendbar ist und die einzige Anforderung ist, Patientendaten vor der Angiographie aufzunehmen. Die komplette Methode ist vollständig automatisch und benötigt zwei Minuten zur Segmentierung und zum Vorschlagen eines Katheters.