Visual Exploration of Image Archives and Datasets
Integrating web visualization technology, image archive, and numerical analysis tools, the Imaging group at ABCC has developed a prototyping system to help navigation through large size data sets derived from image analysis and experiments. The concept of visual exploration gives researchers direct interaction with images and scientific data stored hierarchically in NoSQL (not only SQL) databases using searching, comparison, and numerical analysis capabilities. The result is a web-based, zero-footprint visualization system for image archives and datasets.
Vascular Network Quantification on Angiography Images
An analysis framework has been developed to assess the quality and accuracy of vessel segmentation algorithms for three dimensional images. We generate synthetic (in silico) vessel models which act as ground truth and are constructed to embody varying morphological features. These models are transformed into images constructed under different levels of contrast, noise, and intensity. To demonstrate the use of our framework, we implemented two segmentation algorithms and compare the results to the ground truth model using several measures to quantify the accuracy and quality of segmentation. Furthermore, we collect metrics which describe the characteristics of the vessels it fails to segment. Our approach is illustrated with several examples.
Link to the publication.
Quick2Insight: A User-Friendly Framework for Interactive Rendering of Biological Image Volumes
A new framework has been developed for simple, interactive volume exploration of biological datasets. We accomplish this by automatically creating dataset-specific transfer functions and utilizing them during direct volume rendering. Our method employs a K-Means++ clustering algorithm to classify a two-dimensional histogram created from the input volume. The classification process utilizes spatial and data properties from the volume. Then using properties derived from the classified clusters, our method automatically generates color and opacity transfer functions and presents the user with a high quality initial rendering of the volume data. Our method estimates classification parameters automatically, yet users are also allowed to input or override parameters to utilize pre-existing knowledge of their input data. User input is incorporated through the simple yet intuitive interface for transfer function manipulation included in our framework. Our new interface helps users focus on feature space exploration instead of the usual effort intensive, low-level widget manipulation. We evaluated the framework using three-dimensional medical and biological images. Our preliminary results demonstrate the effectiveness of our method of automating transfer function generation for high quality initial visualization. Our approach effectively generates automatic transfer functions and enables users to explore and interact with their data in an intuitive way, without requiring detailed knowledge of computer graphics or rendering techniques.
Link to the publication.
Bring Dynamic Programming Segmentation Framework into Insight Toolkit
The dynamic programming segmentation algorithm the framework implemented was previously published at NCI-Frederick. The segmentation framework developed integrates the dynamic programming algorithm into ITK and Slicer, a popular medical image viewer, which gives a uniform interface for segmentation by the expert users.
Quantitative Analysis Workflow for Metastasis Study
Collaborated with the SAIP, the Imaging group has developed automatic metastasis study workflow for quick segmentation and measurement of metastatic tumors. The workflow enables researchers to batch process MRI metastasis images using consistent parameter sets. The result is fast measurement of metastatic tumors free of human subjective errors.