Novel Reconstruction Paradigm for Multiphasic CT Imaging of Kidney Cancer

University Of Utah, Salt Lake City UT

Investigators

Linked publications & trials

Abstract

A multiphase CT scan acquires multiple CT images separated by short time intervals in order to capture different contrast enhancement patterns. These different patterns provide complementary information for diagnosis and disease staging. Multiphase CT scans are common in abdominal and pelvis exams. In particular, indeterminate renal lesions or hematuria require four CT acquisitions to distinguish all renal lesion subtypes. The benefit of multiphase CT is unequivocal. However, a four-phase CT scan has on average four times the radiation dose of a single CT scan. Moreover, these exams are often performed periodically to monitor disease progression, which raises serious concerns in terms of risks associated with radiation dose imparted to the patient. The CT images in a multiphase scan differ from each other in terms of contrast enhancement patterns, and also slightly in terms of organ position due to patient breathing. Nevertheless, there is a high level of spatial and intensity correlation between these CT images. Current clinical reconstruction methods reconstruct each phase of a multiphase CT exam independently, discarding the correlation. The correlated image content of all CT phases is currently an untapped source of information that will enable dramatic improvements in image quality and dose utilization. In this application, we aim to develop a novel image reconstruction paradigm that jointly reconstructs all images in a multiphase scan. In this paradigm, the image at each individual phase will benefit from all acquired data in a multiphase acquisition. The improved image quality can then be used to reduce radiation dose to the patient, or to more confidently and perhaps more frequently follow the response of lesions to therapies, both standard of care and investigational. The proposed algorithm builds on state-of-the-art model based iterative reconstruction methods; it utilizes self-similarity both within a single image and among the multiple correlated images in a multiphase scan. Objective, task-based assessment of image quality (IQ) is critical to the success of this research proposal. We will use this type of assessment during algorithm development for progress monitoring and guidance, and also to carefully evaluate image quality gains between the proposed and competing methods as achieved after 18 months of development. The IQ evaluation includes two components: noise reduction which is directly linked to dose reduction, and anatomical accuracy. We will design lesion detection and anatomical error detection tasks, and rely on task-specific metrics associated with receiver operating characteristic methodology, using both model and human observers. The expected outcome of the proposal is a software prototype that is applicable to real patient data with initial clinically-relevant proof of improved image quality, which could be traded for lower radiation dose. The targeted reduction in dose is a factor close to the number of phases. The proposal will enable a paradigm shift in data processing and analysis for multiphase CT scans of kidney cancer patients.