Title:	HUDI: Huge Complex Diagnostic Imaging Data
DNr:	Berzelius-2024-373
Project Type:	LiU Berzelius
Principal Investigator:	Jonas Lantz <jonas.lantz@liu.se>
Affiliation:	Linköpings universitet
Duration:	2024-10-01 – 2025-04-01
Classification:	20603
Homepage:	https://www.liu.se/cmr
Keywords:

Abstract

Personalized medicine is at the center of all discussions of future medicine. The promise is to enable earlier diagnoses, better risk assessments, and optimal treatments by providing patient-specific care and treatment. The foundation for this is the generation of realistic patient models which are based to a large extent on high-quality imaging modalities. The goal of this project is to provide a pipeline for the generation of personalized heart models. This comprises collecting data using the image acquisition technology of the future and the development of methods that support an efficient and effective analysis of the data for diagnosis and model building. Thereby the focus will be on high-resolution spectral data from photon-counting detector Computer Tomography (PCD-CT) imaging of the heart as the basis for the reconstruction of time-varying data representing the heart anatomy, its dynamics, the blood flow, and myocardial strain providing insights into the functioning of the heart. Computer-Tomography (CT) has traditionally been the imaging modality creating the largest data volumes per exam. However, the coming generation CT, photon-counting detector CT (PCD-CT), is not a normal improvement, but a quantum step in information, data quality and data volume, which requires major improvements throughout the whole data handling pipeline. PCD-CT allows for at least a doubling the resolution, resulting in 8 times the data volume of traditional CT, while the spectral data increases the data volume 8 times as well. The quality of data also facilitates the extraction of new information as regional assessment of the strain tensor of the heart muscle, making the data handling even more challenging. Starting from traditional CT an efficient segmentation method for the major anatomical structures of the heart will be developed using Deep Learning. Utilizing the models created using traditional CT a high-resolution segmentation of the heart walls and valves will be developed for PCD-CT. Based on the segmented model, functional data will be extracted. The displacement in the heart muscle will be tracked regionally to obtain the strain tensor in the muscle. The blood flow in the blood pool will be obtained through CFD simulation, using the patient-specific boundary conditions from the segmentations. With this new method we propose to provide additional information about the patients’ unique cardiac physiology and function