Frederik Temmermans

3D human pose recognition for home monitoring of elderly

A toolbox for the automatic monitoring of elderly in a nursing home or in the natural home environment is proposed. Rather than monitoring vital signs or other biomedical parameters, the toolbox is focussed on the monitoring of activity patterns and changes therein. Activity information is derived from visual information using image processing algorithms. The visual information is acquired using 3D camera technology. Besides a traditional visual image, 3D cameras also provide highly accurate depth information. The 3D position of the subject is derived and serves as the primary information source for the different components in the toolbox.

JPEG's JPSearch Standard: Harmonizing Image Management and Search

Triggered by the rise of social networks, community-based image sharing platforms are emerging at an increasing rate. Currently, almost every repository offers a different interaction interface and metadata description format. Unfortunately, this prevents unified and efficient access to these repositories. Consequently, data exchange between systems is often cumbersome. In this context, ISO/IEC JTC1 SC29 WG1 (more commonly known as JPEG) initiated the JPSearch framework standardization, which aims to foster the interaction with and among image repositories. The standard focuses on three main cornerstones supporting repository synchronization, search and access, and image collection creation and maintenance. This article discusses the main concepts, parts, and achievements of the JPSearch framework and demonstrates its use through a set of substantial case studies.

JPSearch: An answer to the lack of standardization in mobile image retrieval

The amount of imagery that is available via various content providers grows at a staggering rate. Contemporary handheld devices allow accessing these content providers anytime, anyplace. Moreover, using the build-in camera, one can easily contribute to these online image repositories. The present lack of consistency in how repositories are accessed for retrieval or syncing complicates inter-operability between devices and systems. JPSearch, a recent standardization initiative within the Joint Photographic Experts Group (JPEG) committee, resolves this predicament by defining the interfaces and protocols for data exchange between them, while restricting as little as possible how those devices, systems or components perform their task.

JPSearch: Metadata Interoperability During Image Exchange [Standards in a Nutshell]

Billions of new images are generated daily. However, the management and exchange of the associated metadata remains cumbersome. Therefore, the Joint Photographic Experts Group (JPEG) has recently promoted a new set of technologies enabling the interoperability between image repositories and/or their clients to an international standard. JPSearch is a suite of specifications that supports the enrichment with metadata data stored in JPEG or JPEG 2000 image formats. It addresses schema and ontology building blocks, a query format, a file format for metadata embedded in image data, and a data interchange format for image repositories.

Lesion segmentation algorithm for contrast enhanced CT images

Lesion segmentation is crucial in liver pathology diagnosis and surgery planning. Its goal is to identify lesions shape, location and connectivity to the vessel system. Nowadays, the clinical practice is to use volumetric contrast enhanced CT images, acquired before and after contrast agent injection. However, currently, liver CAD systems work on a single volumetric image, i.e. the volume exhibiting the best contrast enhancement. Therefore, the motivation of our work is to explore the gray-level enhancement in the different abdominal tissues and organs present in all acquired images. The described method at first aligns all images and progresses with the segmentation - a combination of an initial clustering approach and the Expectation-Maximization algorithm to optimally model the joint histogram by a sum of multivariate Gaussian distributions. It is performed hierarchically: firstly, it is invoked on the whole abdominal volume, secondly on the detected liver region, and finally over the lesions. Experiments show that if the contrast information is sufficient, the results are accurate in comparison to the ground truth: reference segmentation performed by a radiologist using a commercial tool. Moreover, we show that our method provides beneficial information to radiologists about the lesion nature and its behavior throughout the different phases.

Automated Estimation of Hip Prosthesis Migration: a Feasibility Study

A common complication associated with hip arthoplasty is prosthesis migration, and for most cemented components a migration greater than 0.85 mm within the first six months after surgery, are an indicator for prosthesis failure. Currently, prosthesis migration is evaluated using X-ray images, which can only reliably estimate migrations larger than 5 mm. We propose an automated method for estimating prosthesis migration more accurately, using CT images and image registration techniques. We report on the results obtained using an experimental set-up, in which a metal prosthesis can be translated and rotated with respect to a cadaver femur, over distances and angles applied using a combination of positioning stages. Images are first preprocessed to reduce artefacts. Bone and prosthesis are extracted using consecutive thresholding and morphological operations. Two registrations are performed, one aligning the bones and the other aligning the prostheses. The migration is estimated as the difference between the found transformations. We use a robust, multi-resolution, stochastic optimization approach, and compare the mean squared intensity differences (MS) to mutual information (MI). 30 high-resolution helical CT scans were acquired for prosthesis translations ranging from 0.05 mm to 4 mm, and rotations ranging from 0.3° to 3° . For the translations, the mean 3D registration error was found to be 0.22 mm for MS, and 0.15 mm for MI. For the rotations, the standard deviation of the estimation error was 0.18° for MS, and 0.08° for MI. The results show that the proposed approach is feasible and that clinically acceptable accuracies can be obtained. Clinical validation studies on patient images will now be undertaken.

Classification of microcalcifications using micro-CT

Microcalcifications are tiny spots of calcium deposit that often occur in female breasts. Microcalcifications are common in healthy woman, but they often are an early sign of breast cancer. On a mammogram; the current standard of care for breast screening; calcifications appear as tiny white dots. They may occur scattered throughout the breast or grouped in clusters. Radiologists determine the suspiciousness based upon several factors, including position, frequency, grouping, evolution compared to prior studies and shape. In this paper, we study micro-CT images of biopsy samples containing microcalcifications. The scanner delivers 3D images with a voxel size of 8.66 μm, i.e. ca. 8 times the spatial resolution of a contemporary digital mammogram. We propose an automated binary classification method of the samples, based upon shape analysis of the microcalcifications. The study is performed on a set of 50 benign and 50 malign samples preserved in paraffin. The ground truth of the classification is based upon anapathological investigation of the paraffin blocks. The results show a sensitivity, i.e. the percentage of correctly classified malign samples, of up to 98% with a specificity of 40%.

Liver and Lesion Segmentation Algorithm for Contrast Enhanced CT Images

Automatic liver segmentation is a crucial step for aiding in liver surgery and in diagnosing liver pathologies. Its goal is to identify the important anatomical structures as the liver segments, the vessel tree, present lesion and tumors. Nowadays, the current clinical practice is to use volumetric dynamic contrast enhanced computer tomography images, acquired before and after contrast agent injection. However, currently, liver computer aided diagnosis systems work with a single volumetric image, i.e. the volume exhibiting the best contrast enhancement. Therefore, the motivation of our work is to explore the gray-level enhancement in the different abdominal tissues and organs present in all acquired volumes (phases). The described method first brings into alignment all volumes. The segmentation combines an initial clustering approach with the EM algorithm to optimally fit a sum of multivariate Gaussian distributions to the multidimensional joint histogram. The segmentation is performed in an hierarchical way: the whole abdominal volume is segmented in the first step, the segmentation is applied only on the detected liver region in the second step and only at the individual lesions in the third step. We give a comparison of the segmentation results to the ground truth data obtained via manual segmentation by a radiologist.

Shape-Based Clustering and Classification of Breast Microcalcifications in Micro-CT Images

This paper presents a novel classification approach for microcalcifications extracted from core biopsy tissue samples digitized using micro-CT, a high-resolution 3D imaging modality. Microcalcifications are tiny spots of calcium that may occur in the female breast. Although they are common in healthy woman, they are often an early sign of breast cancer. In case of suspiciousness, a biopsy is conducted and the extracted tissue is pathologically analyzed for the presence of cancer cells. The calcifications themselves however are mostly not analyzed. As a result, there is a ground truth for the tissue samples but not for the individual calcifications. Therefore, this paper presents an approach where the calcifications are first classified according to shape-based clusters before predicting a label for the tissue samples.

Towards standardized integration of images in the cloud of linked data

Currently, there are several ways of describing and referring to images in RDF. This ambiguity results in a proliferation of vocabularies for image descriptions, complicating the cross-community data integration of information related to digital images. In addition, there are no standardized guidelines on how to integrate RDF data into the image metadata. Therefore, the JPEG standardization committee has recently initiated some activities to streamline the integration of images in the cloud of Linked Data. One effort is the standardization of an ontology for describing digital images. JPEG aims at providing a technology that enables the uniform description of photos and videos with technical, administrative and semantic metadata compliant with the RDF specification and the principles of Linked Data. Secondly, specifications to integrate RDF data into JPEG images are elaborated. Finally, since descriptions often only apply to a certain part of an image, a last effort is to formalize the specification of regions of interest. In this paper, members of the JPEG committee provide a detailed overview of these ongoing activities.