Search

Accurate localization of anatomical landmarks is an important step in medical imaging, as it provides useful prior information for subsequent image analysis and acquisition methods. It is particularly useful for initialization of automatic image analysis tools (e.g. segmentation and registration) and detection of scan planes for automated image acquisition. Landmark localization has been commonly performed using learning based approaches, such as classifier and/or regressor models. However, trained models may not generalize well in heterogeneous datasets when the images contain large differences due to size, pose and shape variations of organs. To learn more data-adaptive and patient specific models, we propose a novel stratification based training model, and demonstrate its use in a decision forest. The proposed approach does not require any additional training information compared to the standard model training procedure and can be easily integrated into any decision tree framework. The proposed method is evaluated on 1080 3D highresolution and 90 multi-stack 2D cardiac cine MR images. The experiments show that the proposed method achieves state-of-theart landmark localization accuracy and outperforms standard regression and classification based approaches. Additionally, the proposed method is used in a multi-atlas segmentation to create a fully automatic segmentation pipeline, and the results show that it achieves state-of-the-art segmentation accuracy.

Quantification of left (LV) and right (RV) ventricular volumes and masses from Cardiac Magnetic Resonance (CMR) images is of prime importance for the clinical assessment of a wide variety of cardiac diseases. Despite over a decade of research aimed at the development of fast and reliable tools for automated endo- and epicardial contours identification, the problem is still open, particularly for the RV as a consequence of its more irregular shape and its higher density of trabeculations. In this study, a novel near-automated technique for the segmentation of LV endo- and epicardial as well as RV endocardial contours is presented. The technique is based on a 3D narrow-band statistical level set and on 2D edge-based level set algorithms. The technique was tested on CMR images acquired at both end-diastolic and end-systolic phases. For performance evaluation, an experienced interpreter manually traced ventricular contours, which were used as reference. A series of quantitative error metrics (e.g. mean absolute distance, MAD) were computed between automatically identified and manually traced contours. The results showed the high accuracy of the proposed technique (MAD: LV Endo = 1.4±0.7 px; RV Endo = 1.6±1.2 px; LV Epi = 1.4±0.6 px), which could thus potentially lead to the implementation of a tool for fast and reliable identification of ventricular contours.