Petia Radeva

Discriminant ECOC: a heuristic method for application dependent design of error correcting output codes

We present a heuristic method for learning error correcting output codes matrices based on a hierarchical partition of the class space that maximizes a discriminative criterion. To achieve this goal, the optimal codeword separation is sacrificed in favor of a maximum class discrimination in the partitions. The creation of the hierarchical partition set is performed using a binary tree. As a result, a compact matrix with high discrimination power is obtained. Our method is validated using the UCI database and applied to a real problem, the classification of traffic sign images.

Traffic Sign Recognition Using Evolutionary Adaboost Detection and Forest-ECOC Classification

The high variability of sign appearance in uncontrolled environments has made the detection and classification of road signs a challenging problem in computer vision. In this paper, we introduce a novel approach for the detection and classification of traffic signs. Detection is based on a boosted detectors cascade, trained with a novel evolutionary version of Adaboost, which allows the use of large feature spaces. Classification is defined as a multiclass categorization problem. A battery of classifiers is trained to split classes in an Error-Correcting Output Code (ECOC) framework. We propose an ECOC design through a forest of optimal tree structures that are embedded in the ECOC matrix. The novel system offers high performance and better accuracy than the state-of-the-art strategies and is potentially better in terms of noise, affine deformation, partial occlusions, and reduced illumination.

On the Decoding Process in Ternary Error-Correcting Output Codes

A common way to model multiclass classification problems is to design a set of binary classifiers and to combine them. Error-correcting output codes (ECOC) represent a successful framework to deal with these type of problems. Recent works in the ECOC framework showed significant performance improvements by means of new problem-dependent designs based on the ternary ECOC framework. The ternary framework contains a larger set of binary problems because of the use of a ldquodo not carerdquo symbol that allows us to ignore some classes by a given classifier. However, there are no proper studies that analyze the effect of the new symbol at the decoding step. In this paper, we present a taxonomy that embeds all binary and ternary ECOC decoding strategies into four groups. We show that the zero symbol introduces two kinds of biases that require redefinition of the decoding design. A new type of decoding measure is proposed, and two novel decoding strategies are defined. We evaluate the state-of-the-art coding and decoding strategies over a set of UCI machine learning repository data sets and into a real traffic sign categorization problem. The experimental results show that, following the new decoding strategies, the performance of the ECOC design is significantly improved.

Human activity recognition from accelerometer data using a wearable device

Activity Recognition is an emerging field of research, born from the larger fields of ubiquitous computing, context-aware computing and multimedia. Recently, recognizing everyday life activities becomes one of the challenges for pervasive computing. In our work, we developed a novel wearable system easy to use and comfortable to bring. Our wearable system is based on a new set of 20 computationally efficient features and the Random Forest classifier. We obtain very encouraging results with classification accuracy of human activities recognition of up to 94%.

Subclass Problem-Dependent Design for Error-Correcting Output Codes

A common way to model multiclass classification problems is by means of Error-Correcting Output Codes (ECOCs). Given a multiclass problem, the ECOC technique designs a code word for each class, where each position of the code identifies the membership of the class for a given binary problem. A classification decision is obtained by assigning the label of the class with the closest code. One of the main requirements of the ECOC design is that the base classifier is capable of splitting each subgroup of classes from each binary problem. However, we cannot guarantee that a linear classifier model convex regions. Furthermore, nonlinear classifiers also fail to manage some type of surfaces. In this paper, we present a novel strategy to model multiclass classification problems using subclass information in the ECOC framework. Complex problems are solved by splitting the original set of classes into subclasses and embedding the binary problems in a problem-dependent ECOC design. Experimental results show that the proposed splitting procedure yields a better performance when the class overlap or the distribution of the training objects conceal the decision boundaries for the base classifier. The results are even more significant when one has a sufficiently large training size.

Deformable B-Solids and Implicit Snakes for 3D Localization and Tracking of SPAMM MRI Data

To date, MRI?SPAMM data from different image slices have been analyzed independently. In this paper, we propose an approach for 3D tag localization and tracking of SPAMM data by a novel deformable B-solid. The solid is defined in terms of a 3D tensor product B-spline. The isoparametric curves of the B-spline solid have special importance. These are termed implicit snakes as they deform under image forces from tag lines in different image slices. The localization and tracking of tag lines is performed under constraints of continuity and smoothness of the B-solid. To track motion from boundaries and restrict image forces to the myocardium, a volumetric model is employed as a pair of coupled endocardial and epicardial B-spline surfaces. To recover deformations of the left ventricle (LV), an energy-minimization problem is posed where both tag and LV boundary data are used. The framework has been implemented on tag data from short axis (SA) cardiac images, as well as SA LV boundaries, and is currently being extended to include long axis data.

Late Stent Recoil of the Bioabsorbable Everolimus-Eluting Coronary Stent and its Relationship With Plaque Morphology

OBJECTIVES This study sought to evaluate late recoil of a novel bioabsorbable everolimus-eluting coronary stent (BVS), which is composed of a poly-L-lactic acid backbone, coated with a bioabsorbable polymer containing everolimus. BACKGROUND Little is known about the mechanical behavior of bioabsorbable polymer stents after deployment in diseased human coronary arteries. METHODS The study population consisted of 16 patients, who were treated with elective BVS implantation for single de novo native coronary artery lesions and were followed at 6 months. All patients underwent an intravascular ultrasound examination at post-procedure and follow-up. A total of 484 paired cross-sectional areas (CSAs) were acquired and analyzed. Late absolute stent recoil was defined as stent area at post-procedure (X) - stent area at follow-up (Y). Late percent stent recoil was defined as (X - Y)/X x 100. In each CSA, plaque morphology was assessed qualitatively and classified as calcific, fibronecrotic, or fibrocellular plaque. RESULTS Late absolute and percent recoil of the BVS was 0.65 +/- 1.71 mm(2) (95% confidence interval [CI]: 0.49 to 0.80 mm(2)) and 7.60 +/- 23.3% (95% CI: 5.52% to 9.68%). Calcified plaques resulted in significantly less late recoil (0.20 +/- 1.54 mm(2) and 1.97 +/- 22.2%) than fibronecrotic plaques (1.03 +/- 2.12 mm(2) and 12.4 +/- 28.0%, p = 0.001 and p = 0.001, respectively) or fibrocellular plaque (0.74 +/- 1.48 mm(2) and 8.90 +/- 19.8%, p = 0.001 and p = 0.001, respectively). CONCLUSIONS The BVS shrank in size during the follow-up period. The lesion morphology of stented segments might affect the degree of late recoil of the BVS. (ABSORB Everolimus Eluting Coronary Stent System First in Man Clinical Investigation; NCT00300131).

Tag surface reconstruction and tracking of myocardial beads from SPAMM-MRI with parametric B-spline surfaces

Magnetic resonance imaging (MRI) is unique in its ability to noninvasively and selectively alter tissue magnetization, and create tag planes intersecting image slices. The resulting grid of signal voids allows for tracking deformations of tissues in otherwise homogeneous-signal myocardial regions. Here, the authors propose a specific spatial modulation of magnetization (SPAMM) imaging protocol together with efficient techniques for measurement of three-dimensional (3-D) motion of material points of the human heart (referred to as myocardial beads) from images collected with the SPAMM method. The techniques make use of tagged images in orthogonal views by explicitly reconstructing 3-D B-spline surface representation of tag planes (tag planes in two orthogonal orientations intersecting the short-axis (SA) image slices and tag planes in an orientation orthogonal to the short-axis tag planes intersecting long-axis (LA) image slices). The developed methods allow for viewing deformations of 3-D tag surfaces, spatial correspondence of long-axis and short-axis image slice and tag positions, as well as nonrigld movement of myocardial beads as a function of time.

Automatic Detection of Intestinal Juices in Wireless Capsule Video Endoscopy

Wireless capsule video endoscopy is a novel and challenging clinical technique, whose major reported drawback relates to the high amount of time needed for video visualization. In this paper, we propose a method for the rejection of the parts of the video resulting not valid for analysis by means of automatic detection of intestinal juices. We applied Gabor filters for the characterization of the bubble-like shape of intestinal juices in fasting patients. Our method achieves a significant reduction in visualization time, with no relevant loss of valid frames. The proposed approach is easily extensible to other image analysis scenarios where the described pattern of bubbles can be found

New insight into intestinal motor function via noninvasive endoluminal image analysis.

BACKGROUND & AIMS Evaluation of small bowel motility by intestinal manometry is invasive and requires expertise for interpretation. Our aim was to use capsule technology for evaluation of small bowel motor function based on a fully computerized image analysis program. METHODS Thirty-six consecutive patients with severe intestinal motor disorders (19 fulfilling manometric criteria of intestinal dysmotility and 17 not) and 50 healthy subjects received the endoscopic capsule (Pillcam; Given Imaging, Yokneam, Israel). Endoluminal image analysis was performed with a computer vision program specifically developed for the detection of contractile patterns (phasic luminal closure and radial wrinkles by wall texture analysis), noncontractile patterns (tunnel and wall appearance by Laplacian filtering), intestinal content (by color decomposition analysis), and endoluminal motion (by chromatic stability). Automatic classification of normal and abnormal intestinal motility was performed by means of a machine-learning technique. RESULTS As compared with healthy subjects, patients exhibited less contractile activity (25% less phasic luminal closures, P < .05) and more noncontractile patterns (151% more tunnel pattern, P < .05), static sequences (56% more static images, P < .01), and turbid intestinal content (94% more static turbid images, P < .01). On cross validation, the classifier identified as abnormal all but 1 patient with manometric criteria of dysmotility and as normal all healthy subjects. Out of the 17 patients without manometric criteria of dysmotility, 11 were identified as abnormal and 6 as normal. CONCLUSIONS Our study shows that endoluminal image analysis, by means of computer vision and machine-learning techniques, constitutes a reliable, noninvasive, and automated diagnostic test of intestinal motor disorders.