Rongxin Li

Towards a Visual Programming Environment Based on ITK for Medical Image Analysis

Medical image analysis experiments usually require a certain level of programming skills and are often time-consuming. Visual programming can address these issues by providing a high-level user interface to visually access and combine underlying medical image analysis functionalities. This paper describes the ITKBoard, an extension of the Insight Toolkit for Registration and Segmentation (ITK). Unlike traditional approaches, ITKBoard, a visual tool for experimentation and prototyping, allows rapid visual construction of filter pipelines, via a graphical user interface. An overview of the main functionalities and the system design is presented in this paper. The approaches that we have developed to resolving critical issues such as pipeline handling, automatic object wrapping and plugin mechanism are discussed in detail. This is a first step towards a visual programming environment designed for medical image analysis, in which the specific needs in terms of visual data-flow and control-flow programming will be progressively addressed. It is envisaged that when the software is released in the near future it will be significantly beneficial to the medical image analysis community.

Combining front propagation with shape knowledge for accurate curvilinear modelling

In this paper, we propose an approach to accurately modelling tubular, anatomical structures as curvilinear entities. Current optimal path and centerline extraction techniques are either prone to introducing spurious tortuosity, or unable to consistently avoid taking shortcuts at high curvature positions. These problems not only affect spatial appreciation of the structure but may also significantly impact the accuracy of length, angle and tortuosity measurements. Our approach overcomes the above deficiencies through the combination of a front propagation method and a model in which a priori shape knowledge is embedded. This approach is designed to be used in endovascular and neurological surgical planning. The efficacy of our method is demonstrated using synthetic and clinical data.

Toward Consistently Behaving Deformable Models For Improved Automation In Image Segmentation

Deformable models are a powerful approach to medical image segmentation. However, currently the behavior of a deformable model is highly dependent on its initialization and parameter settings. This is an obstacle to robust automatic or near-automatic segmentation. A generic approach to reducing this dependency is introduced in the present chapter based on topographic distance transforms from manually or automatically placed markers. This approach utilizes object and background differentiation through watershed theories. The implementation is based on efficient numerical methods such as the Fast Marching method and non-iterative reconstruction-by-erosion. Further extension into a multi-region coupled segmentation approach is discussed. Validation experiments are presented to demonstrate the capabilities of this approach. A preliminary application in pediatric dosimetry research is described. It is believed that the more consistent behavior will enable a higher degree of automation for segmentation employing deformable models and is particularly suited for applications that involve segmentation-based construction of organ models from image databases, especially in situations where the markers can be placed automatically based on a priori knowledge.

A marker-induced vector field for reduced sensitivity to initialization for parametric and geometric deformable models

Deformable models are powerful approaches to medical image analysis, particularly segmentation. However, the outcome of applying a parametric or geometric deformable model is often significantly dependent on its initialization. This is an obstacle to robust automatic segmentation. Based on theoretical analyses of the watershed transform, we propose a novel approach to reducing this sensitivity to initialization by deriving a vector field from topographic and Euclidean distance transforms. This vector field is aimed to extend the influence of the gradients at the boundary of the segmentation target over the entire image in a consistent fashion, while ignoring any irrelevant gradients in the original image. Initiated by one or more segmentation seeds, the vector field is first computed using an efficient numerical method, and subsequently participates in the models evolution process. Integration of the vector field has so far been performed with a two-dimensional (2D) parametric deformable model and with a three-dimensional (3D) geodesic active contour level set model. We believe that our approach will enable a higher degree of automation for deformable-model-based segmentation, particularly in situations where the seeds can be placed automatically based on, for example, a priori knowledge regarding the anatomy and the intensity differentiation between the target and the background. Experiments on segmenting organs and tumors from CT and MR images using the integrated models have shown that this is a promising approach.

A Visual Dataflow Language for Image Segmentation and Registration

Experimenters in biomedical image processing rely on software libraries to provide a large number of standard filtering and image handling algorithms. The Insight Toolkit (ITK) is an open-source library that provides a complete framework for a range of image processing tasks, and is specifically aimed at segmentation and registration tasks for both two and three dimensional images.

Using marker-based deformable models for deriving anatomical models from medical image databases

The outcome of the application of a deformable model is dependent on its initialization and parameter settings. This is an obstacle to robust automatic segmentation. An approach to reducing the dependences is proposed in this paper based on geodesic topographic distance transforms from two manually or automatically placed markers. This is a novel method that uses object and background differentiation in deformable models via watershed theories. The implementation is based on efficient numerical methods. Experiments have shown that it is a highly promising approach. This approach has been applied to the determination of tissue characteristics in early childhood using a CT database of neonates. We believe that by relaxing dependences on the initialization and parameters this approach will enable a degree of automation needed for segmentation-based construction of organ models from large image databases, particularly when the markers can be determined automatically on the basis of a priori knowledge