Search

Determining spinal geometry and in particular the position and orientation of the intervertebral disks is an integral part of nearly every spinal examination with Computed Tomography (CT) and Magnetic Resonance (MR) imaging. It is particularly important for the standardized alignment of the scan geometry with the spine. In this paper, we present a novel method that combines Marginal Space Learning (MSL), a recently introduced concept for efficient discriminative object detection, with a generative anatomical network that incorporates relative pose information for the detection of multiple objects. It is used to simultaneously detect and label the intervertebral disks in a given spinal image volume. While a novel iterative version of MSL is used to quickly generate candidate detections comprising position, orientation, and scale of the disks with high sensitivity, the anatomical network selects the most likely candidates using a learned prior on the individual nine dimensional transformation spaces. Since the proposed approach is learning-based it can be trained for MR or CT alike. Experimental results based on 42 MR volumes show that our system not only achieves superior accuracy but also is the fastest system of its kind in the literature - on average, the spinal disks of a whole spine are detected in 11.5s with 98.6% sensitivity and 0.073 false positive detections per volume. An average position error of 2.4mm and angular error of 3.9° is achieved.

In his original paper on random forests, Breiman proposed two different decision tree ensembles: one generated from "orthogonal" trees with thresholds on individual features in every split, and one from "oblique" trees separating the feature space by randomly oriented hyperplanes. In spite of a rising interest in the random forest framework, however, ensembles built from orthogonal trees (RF) have gained most, if not all, attention so far. In the present work we propose to employ "oblique" random forests (oRF) built from multivariate trees which explicitly learn optimal split directions at internal nodes using linear discriminative models, rather than using random coefficients as the original oRF. This oRF outperforms RF, as well as other classifiers, on nearly all data sets but those with discrete factorial features. Learned node models perform distinctively better than random splits. An oRF feature importance score shows to be preferable over standard RF feature importance scores such as Gini or permutation importance. The topology of the oRF decision space appears to be smoother and better adapted to the data, resulting in improved generalization performance. Overall, the oRF propose here may be preferred over standard RF on most learning tasks involving numerical and spectral data.

Recently conducted clinical studies prove the utility of Coronary Computed Tomography Angiography (CCTA) as a viable alternative to invasive angiography for the detection of Coronary Artery Disease (CAD). This has lead to the development of several algorithms for automatic detection and grading of coronary stenoses. However, most of these methods focus on detecting calcified plaques only. A few methods that can also detect and grade non-calcified plaques require substantial user involvement. In this paper, we propose a fast and fully automatic system that is capable of detecting, grading and classifying coronary stenoses in CCTA caused by all types of plaques. We propose a four-step approach including a learning-based centerline verification step and a lumen crosssection estimation step using random regression forests.We show state-of-the-art performance of our method in experiments conducted on a set of 229 CCTA volumes. With an average processing time of 1.8 seconds per case after centerline extraction, our method is significantly faster than competing approaches.

We present a scheme for the development of a spot detection procedure which is based on the learning of latent linear features from a training data set. Adapting ideas from face recognition to this low level feature extraction task, we suggest to learn a collection of filters from representative data that span a subspace which allows for a reliable distinction of a spot vs. the heterogeneous background; and to use a non-linear classifier for the actual decision. Comparing different subspace projections, in particular principal component analysis, partial least squares, and linear discriminant analysis, in conjunction with subsequent classification by random forests on a data set from archaeological remote sensing, we observe a superior performance of the subspace approaches, both compared with a standard template matching and a direct classification of local image patches.