Learning-Based Detection and Tracking in Medical Imaging: A Probabilistic Approach

Abstract

Medical image processing tools are playing an increasingly important role in assisting the clinicians in diagnosis, therapy planning and image-guided interventions. Accurate, robust and fast tracking of deformable anatomical objects, such as the heart, is a crucial task in medical image analysis. One of the main challenges is to maintain an anatomically consistent representation of target appearance that is robust enough to cope with inherent changes due to target movement, imaging device movement, varying imaging conditions, and is consistent with the domain expert clinical knowledge. To address these challenges, this chapter presents a probabilistic framework that relies on anatomically indexed component-based object models which integrate several sources of information to determine the temporal trajectory of the deformable target. Large annotated imaging databases are exploited to encode the domain knowledge in shape models and motion models and to learn discriminative image classifiers for the target appearance. The framework robustly fuses the prior information with traditional tracking approaches based on template match-

Multifidelity-CMA: a multifidelity approach for efficient personalisation of 3D cardiac electromechanical models

“…We finally present results on a large database of 121 cases. For each patient, a biventricular heart mesh geometry (between 10 000 and 15 000 nodes) was built from the available MRI image and the boundaries of the myocardium were tracked in the cine MRI images as described in (Jolly et al, 2011) and (Wang et al, 2013). This led to the computation of the volume curve, then the value of the stroke volume.…”

SVF-Net: Learning Deformable Image Registration Using Shape Matching

“…Our dataset consists of N = 187 short-axis acquisitions of end-diastolic cardiac images acquired in multiple clinical centers (University College London Hospitals, Ospedale Pediatrico Bambino Gesú Roma and Deutsches Herzzentrum Berlin). For each of these images, the myocardium was segmented based on a data-driven approach combining the methods of [6,10] and quality controlled by experts. As a preprocessing step, and to study all the data in a common space, we first rigidly align the images (using the information from the ground truth segmentations) and resample them to have a consistent image size of 64 × 64 × 16 through all the dataset.…”

Automatic image-to-model framework for patient-specific electromechanical modeling of the heart

“…MRI, 3D US, CT or C-arm CT). To that end, we employ a robust, data-driven machine learn- ing approach [8] in order to estimate meshes of the endocardia and epicardium automatically. Appending them yields a closed surface of the biventricular myocardium.…”