Tatjana Zrimec

3D modelling and visualization of the human lung

A method for modelling and visualizing human lungs using knowledge of lung anatomy and high resolution CT (HRCT) images is presented. The model consists of a symbolic description of lung anatomy and a 3D atlas. The 3D atlas is constructed using HRCT volume data. A few anatomical landmarks are determined and are used to divide the lungs into anatomically and diagnostically important regions. The landmarks and the lung regions enable accurate mapping of the model to patient data and enable the system to deal with image and human variability. The model can be displayed as a set of labelled axial slices and as a 3D model of the lungs. The 3D visualization enables rotation and viewing of lung structures, lung features and lung regions from different angles.

3D Lung Image Retrieval Using Localized Features

The interpretation of high-resolution computed tomography (HRCT) images of the chest showing disorders of the lung tissue associated with interstitial lung diseases (ILDs) is time-consuming and requires experience. Whereas automatic detection and quantification of the lung tissue patterns showed promising results in several studies, its aid for the clinicians is limited to the challenge of image interpretation, letting the radiologists with the problem of the final histological diagnosis. Complementary to lung tissue categorization, providing visually similar cases using content-based image retrieval (CBIR) is in line with the clinical workflow of the radiologists. In a preliminary study, a Euclidean distance based on volume percentages of five lung tissue types was used as inter-case distance for CBIR. The latter showed the feasibility of retrieving similar histological diagnoses of ILD based on visual content, although no localization information was used for CBIR. However, to retrieve and show similar images with pathology appearing at a particular lung position was not possible. In this work, a 3D localization system based on lung anatomy is used to localize low-level features used for CBIR. When compared to our previous study, the introduction of localization features allows improving early precision for some histological diagnoses, especially when the region of appearance of lung tissue disorders is important.

Detection of Bronchopulmonary Segments on High-Resolution CT--Preliminary Results

Bronchopulmonary segments are subdivisions of lung lobes and provide detailed description of lung anatomy. They are used in surgical resection planning and airway disease quantification. In this paper, we present a method for determining lung segments in volumetric high-resolution CT (HRCT) using segmental bronchi. The bronchial tree is automatically segmented and manually corrected to ensure optimum accuracy. The bronchopulmonary segments are determined by a 3D volume growing with a novel surface smoothing algorithm. Using the detected bronchoplumonary segments of three normal subjects, we measure the inter-patient variation of position of the segments in three subjects.

A medical image-understanding system

Abstract This paper presents a general methodology for computer interpretation and integration of medical images. An explicit anatomical model, as well as other domain knowledge, is used to facilitate feature extraction and the fusion of images from different modalities. A frame representation is used to implement the description of important anatomical features and other domain knowledge. Frames are also used to implement the control mechanisms that establish communication between the image data and the symbolic knowledge. A prototype system for reconstructing the human cerebral vasculature is presented.

A System for Computer Aided Detection of Diseases Patterns in High Resolution CT Images of the Lungs

Automatic detection of disease patterns in medical images can assist radiologists in image analysis. We present a system for detection of disease patterns demonstrated on HRCT images of the lung. Automated image analysis can be assisted by incorporating into a program information and knowledge that is available to radiologists. Anatomical features and landmarks are first extracted from the images. This information, together with the structure and regions of the lung, that are stored in a model of the lungs, is used in detecting disease patterns. Rules for recognizing different disease patterns are generated using machine learning. The systems performance is demonstrated on detecting two kinds of diseases patterns, one related to structural deformation of the bronchial tree and one showing fibrotic changes of the lung parenchyma. The results show that the system is able to recognize and indicate the existence, size and location of potential lung abnormalities.

A 3D model of the human lung with lung regions characterization

A method for modelling human lungs is presented. The model includes both knowledge of lung anatomy and knowledge of the appearance of objects in high resolutions CT images of the lungs. Symbolic, structural and geometric information in the model is stored in frame structures. Frames allow easy representation of the hierarchical structures that are found in human anatomy. A few anatomical landmarks are determined and used for lung characterization as clinically meaningful regions. The results from automatic landmark segmentation, tested on 1685 images from 84 patient studies, show that the carina, hilum, spinal cord and sternum are quite stable features across patients. The use of anatomical landmarks and lung regions helps the system to deal with image and human variability.

LMIK - learning medical image knowledge: an Internet-based medical image knowledge acquisition framework

As part of the Learning Medical Imaging Knowledge project, we are developing a knowledge-based, machine learning and knowledge acquisition framework for systematic feature extraction and recognition of a range of lung diseases from High Resolution Computed Tomography (HRCT) images. This framework allows radiologists to remotely diagnose and share expert knowledge about lung HRCT interpretation, which is then used to develop a Computer Aided Diagnosis (CAD) system for lung disease. In this paper, we describe the knowledge acquisition system LMIK, which is Internet-based and platform-independent. The LMIK utilises the Internet to provide users with secure access to patient and research data and facilitates communication among highly qualified radiologists and researchers. It is currently used by five radiologists and over 20 researchers and has proved to be an invaluable research tool. Research is underway to develop computer algorithms for automatic diagnosis of lung diseases. In future, these algorithms will be integrated into LMIK to equip it with CAD capabilities to improve diagnostic accuracy of radiologists and extend availability of expert clinical knowledge to wider communities.

A 3D Model of the Human Lung

This paper presents a method for modelling human lungs using knowledge of lung anatomy and High Resolution CT images. The model consists of a symbolic anatomical structure map and an annotated 3D atlas. The model is implemented using Frame structures. Frames provide a good platform for the comprehensive description of anatomical features and for enabling communication between the image data and the symbolic knowledge. A few important landmarks have been determined and used to divide the lung into clinically meaningful regions, which enable accurate mapping of the model to patient data.

Methods for Automatic Honeycombing Detection in HRCT images of the Lung

Honeycombing in High-Resolution CT (HRCT) indicates the presence of a number of diseases involving fibrosis of the lung. Honeycombing is difficult to detect due to its textural and structural appearance, which changes with the progression of the diseases. Structure-based and texture-based methods, developed for detecting the honeycombing pattern, are presented and compared. Machine learning is used to generate rules for honeycomb detection using examples of its appearance in HRCT images, provided by radiologists. The effectiveness of each method was evaluated using cross validation on 16692 examples of regions with and without honeycombing from 42 images of 8 patients.

Mining patterns of dyspepsia symptoms across time points using constraint association rules

In this paper, we develop and implement a framework for constraint-based association rule mining across subgroups in order to help a domain expert find useful patterns in a medical data set that includes temporal data. This work is motivated by the difficulties experienced in the medical domain to identify and track dyspepsia symptom clusters within and across time. Our framework, Apriori with Subgroup and Constraint (ASC), is built on top of the existing Apriori framework. We have identified four different types of phase-wise constraints for subgroups: constraint across subgroups, constraint on subgroup, constraint on pattern content and constraint on rule. ASC has been evaluated in a real-world medical scenario; analysis was conducted with the interaction of a domain expert. Although the framework is evaluated using a data set from the medical domain, it should be general enough to be applicable in other domains.