Advanced image processing technologies and modern image analysis tools are furthering research and development efforts to augment the radiologic interpretation process and improve diagnosis and treatment of diseases. The following are examples of some of the scientific sessions presented during yesterday’s Society for Imaging Informatics in Medicine (SIIM, formerly SCAR) show in Austin, Texas, that focused on image processing.
Reducing false positives for automatic computerized detection of pulmonary embolism
Jianming Liang, PhD, of Siemens Medical Solutions spoke on how capturing the vessel characteristics can be utilized to reduce false positives for automatic detection of pulmonary embolism (PE) in computed tomography angiography (CTA). An early diagnosis to PE is key to survival and CTA has emerged as an accurate diagnostic tool for PE. However, there are hundreds of CT slices in each CTA study. Manual reading is laborious, time consuming and complicated with false positives. A CAD (computer assisted diagnosis) tool can assist radiologists in detecting and characterizing emboli in an accurate, efficient way.
“Advanced image processing tools have helped us develop a new feature-based approach for eliminating false positives in automatic PE detection,” said Liang. “The developed method was fully integrated into our PE CAD system and tested on 72 cases.”
The approach is referred to as a vessel based approach and aims to only segment the relevant areas around the candidate PE locations, thereby avoiding a complete segmentation of the complete pulmonary vessel tree. “We only extract a small vessel segment surrounding the PE candidates, leading to high efficiency,” said Liang.
Leaking-free seeded region growing using branch cut for vessel segmentation in lower extremity CT images
Segmenting vessels in lower extremity CT images is very difficult because of gray level variation, connection to bones and their small sizes, said Dongsung Kim, PhD, of Soongsil University. As a result, researchers have set out to develop a new approach that segments bones and subtracts them from original CT images.
“Our research has focused on segmenting and subtracting bones for vessel segmentation,” said Kim. “We have developed a bone segmentation technique without leakage for detecting junctions and classifying branches by appearance, shape, size change and velocity.”
Kim said that researchers used the Seeded Region Growing (SRG) as its core part, which is one of the fastest segmentation algorithms. Although SRG is fast, it suffers a leakage problem when foreign objects are connected to the target object. “If we can solve the leakage problem with a small amount of computing time, the segmentation method can be sufficiently robust and efficient for routine use,” Kim said.
Kim explained that vessels in lower extremity are necessary to examine for diagnosis of vein anomaly. Although contrast can be used to enhance the vessel structures, it is still difficult to examine vessels because bones such as pelvis, femur, fibula and tibia block some areas of the vessels. Thus vessel segmentation is required.
Relative approach to optical density analysis for quantitative fracture healing assessment
“Are we able to monitor fracture healing,” posed Wojciech Glinkoweski, MD, PhD, of the Medical University of Warsaw. “Yes, however usually combined methods are the best.” And as new treatment methods and enhancing agents are introduced, medical professionals are seeking to develop simple yet sensitive noninvasive methods for quantitative detecting changes in the fracture repair process.
“Available assessment methods (for fracture healing assessment) do exist, such as DEXA scanning, CT and MRI (although both rare due to implants and costs), biomechanical-based methods, ultrasound and ultrasonometry,” said Glinkoweski. However, conventional radiography remains a fundamental and most commonly used method of bone diagnosis.
Researchers at the Medical University of Warsaw developed their own approach to quantitative evaluation of fracture healing, called the RODIA system (relative optical density image analysis), which Glinkoweski presented on during his speech. The technology was created utilizing measure of relative analysis of optical density of digital or digitized radiographs.
“Development of modern image analysis tools may lead to improvement in rationalization of fracture treatment,” said Glinkoweski. “Developed methods may enhance global and