With the advent of computers and natural language processing, it is not surprising to see that humans are trying to use computers to answer questions. By the 1960s, there were systems implemented on the two major models of question answering, IR-based and knowledge-based, to answer questions about sport statistics and scientific facts. This paper reports on the development of a knowledge-based question answering system that is aimed at providing cognitive assistance to radiologists. Our system represents the question as a semantic query to a medical knowledge base. Evidence obtained from textual and imaging data associated with the question is then combined to arrive at an answer. This question answering system has 3 stages: i) question text and answer choices processing, ii) image processing, and iii) reasoning. Currently, the system can answer differential diagnosis and patient management questions, however, we can tackle a wider variety of question types by improving our medical knowledge coverage in the future.
Intravascular ultrasound (IVUS) has been proven a reliable imaging modality that is widely employed in cardiac
interventional procedures. It can provide morphologic as well as pathologic information on the occluded plaques in the
coronary arteries. In this paper, we present a new technique using wavelet packet analysis that differentiates between
blood and non-blood regions on the IVUS images. We utilized the multi-channel texture segmentation algorithm based
on the discrete wavelet packet frames (DWPF). A k-mean clustering algorithm was deployed to partition the extracted
textural features into blood and non-blood in an unsupervised fashion. Finally, the geometric and statistical information
of the segmented regions was used to estimate the closest set of pixels to the lumen border and a spline curve was fitted
to the set. The presented algorithm may be helpful in delineating the lumen border automatically and more reliably prior
to the process of plaque characterization, especially with 40 MHz transducers, where appearance of the red blood cells
renders the border detection more challenging, even manually. Experimental results are shown and they are
quantitatively compared with manually traced borders by an expert. It is concluded that our two dimensional (2-D)
algorithm, which is independent of the cardiac and catheter motions performs well in both in-vivo and in-vitro cases.
Plaque characterization through backscattered intravascular ultrasound (IVUS) signal analysis has been the subject of extensive study for the past several years. A number of algorithms to analyze IVUS images and underlying RF signals to delineate the composition of atherosclerotic plaque have been reported. In this paper, we present several realistic challenges one faces throughout the process of developing such algorithms to characterize tissue type.
The basic tenet of ultrasound tissue characterization is that different tissue types imprint their own "signature" on the backscattered echo returning to the transducer. Tissue characterization is possible to the extent that these echo signals can be received, the signatures read, and uniquely attributed to a tissue type. The principal difficulty in doing tissue characterization is that backscattered RF signals originating as echoes from different groups of cells of the same tissue type exhibit no obvious commonality in appearance in the time domain. This happens even in carefully controlled laboratory experiments.
We describe the method of acquisition and digitization of ultrasound radiofrequency (RF) signals from left anterior descending and left circumflex coronary arteries. The challenge of obtaining corresponding histology images to match to specific regions-of-interest on the images is discussed.
A tissue characterization technique based on seven features is compared to a full spectrum based approach. The same RF and histology data sets were used to evaluate the performances of these two techniques.