Chi-Ren Shyu

ASSERT: a physician-in-the-loop content-based retrieval system for HRCT image databases

It is now recognized in many domains that content-based image retrieval from a database of images cannot be carried out by using completely automated approaches. One such domain is medical radiology for which the clinically useful information in an image typically consists of gray level variations in highly localized regions of the image. Currently, it is not possible to extract these regions by automatic image segmentation techniques. To address this problem, we have implemented a human-in-the-loop (a physician-in-the-loop, more specifically) approach in which the human delineates the pathology bearing regions (PBR) and a set of anatomical landmarks in the image when the image is entered into the database. To the regions thus marked, our approach applies low-level computer vision and image processing algorithms to extract attributes related to the variations in gray scale, texture, shape, etc. In addition, the system records attributes that capture relational information such as the position of a PBR with respect to certain anatomical landmarks. An overall multidimensional index is assigned to each image based on these attribute values.

Relevance feedback decision trees in content-based image retrieval

Significant time and effort has been devoted to finding feature representations of images in databases in order to enable content-based image retrieval (CBIR). Relevance feedback is a mechanism for improving retrieval precision over time by allowing the user to implicitly communicate to the system which of these features are relevant and which are not. We propose a relevance feedback retrieval system that, for each retrieval iteration, learns a decision tree to uncover a common thread between all images marked as relevant. This tree is then used as a model for inferring which of the unseen images the user would not likely desire. We evaluate our approach within the domain of HRCT images of the lung.

Local versus global features for content-based image retrieval

It is now recognized in many domains that content-based image retrieval (CBIR) from a database of images cannot be carried out by using completely automated approaches. One such domain is medical radiology for which the clinically useful information in an image typically consists of gray level variations in highly localized regions of the image. Currently, it is not possible to extract these regions by automatic image segmentation techniques. To address this problem, the authors have implemented a human-in-the-loop (a physician-in-the-loop, more specifically) approach in which the human delineates the pathology bearing regions (PBR) and a set of anatomical landmarks of the image at the time the image is entered into the database. From the regions thus marked, the approach applies low-level computer vision and image processing algorithms to extract features related to the variations of gray scale, texture, shape, etc. The extracted features create an index that characterizes the image. To form an image-based query the physician first marks the PERs. The system then extracts the relevant image features, computes the distance of the query image to all image indices an the database, and retrieves then most similar images. The approach is based on the assumption that medical image characterization must contain features local to the PERs. The focus of the paper is to assess the utility of localized versus global features for the domain of HRCT images of the lung, and to evaluate the systems sensitivity to physician subjectivity in delineating the PBRs.

Entropy-Balanced Bitmap Tree for Shape-Based Object Retrieval From Large-Scale Satellite Imagery Databases

In this paper, we present a novel indexing structure that was developed to efficiently and accurately perform content-based shape retrieval of objects from a large-scale satellite imagery database. Our geospatial information retrieval and indexing system, GeoIRIS, contains 45 GB of high-resolution satellite imagery. Objects of multiple scales are automatically extracted from satellite imagery and then encoded into a bitmap shape representation. This shape encoding compresses the total size of the shape descriptors to approximately 0.34% of the imagery database size. We have developed the entropy-balanced bitmap (EBB) tree, which exploits the probabilistic nature of bit values in automatically derived shape classes. The efficiency of the shape representation coupled with the EBB tree allows us to index approximately 1.3 million objects for fast content-based retrieval of objects by shape.

PhenoPhyte: a flexible affordable method to quantify 2D phenotypes from imagery.

BackgroundAccurate characterization of complex plant phenotypes is critical to assigning biological functions to genes through forward or reverse genetics. It can also be vital in determining the effect of a treatment, genotype, or environmental condition on plant growth or susceptibility to insects or pathogens. Although techniques for characterizing complex phenotypes have been developed, most are not cost effective or are too imprecise or subjective to reliably differentiate subtler differences in complex traits like growth, color change, or disease resistance.ResultsWe designed an inexpensive imaging protocol that facilitates automatic quantification of two-dimensional visual phenotypes using computer vision and image processing algorithms applied to standard digital images. The protocol allows for non-destructive imaging of plants in the laboratory and field and can be used in suboptimal imaging conditions due to automated color and scale normalization. We designed the web-based tool PhenoPhyte for processing images adhering to this protocol and demonstrate its ability to measure a variety of two-dimensional traits (such as growth, leaf area, and herbivory) using images from several species (Arabidopsis thaliana and Brassica rapa). We then provide a more complicated example for measuring disease resistance of Zea mays to Southern Leaf Blight.ConclusionsPhenoPhyte is a new cost-effective web-application for semi-automated quantification of two-dimensional traits from digital imagery using an easy imaging protocol. This tool’s usefulness is demonstrated for a variety of traits in multiple species. We show that digital phenotyping can reduce human subjectivity in trait quantification, thereby increasing accuracy and improving precision, which are crucial for differentiating and quantifying subtle phenotypic variation and understanding gene function and/or treatment effects.

Determining effects of non-synonymous SNPs on protein-protein interactions using supervised and semi-supervised learning.

Single nucleotide polymorphisms (SNPs) are among the most common types of genetic variation in complex genetic disorders. A growing number of studies link the functional role of SNPs with the networks and pathways mediated by the disease-associated genes. For example, many non-synonymous missense SNPs (nsSNPs) have been found near or inside the protein-protein interaction (PPI) interfaces. Determining whether such nsSNP will disrupt or preserve a PPI is a challenging task to address, both experimentally and computationally. Here, we present this task as three related classification problems, and develop a new computational method, called the SNP-IN tool (non-synonymous SNP INteraction effect predictor). Our method predicts the effects of nsSNPs on PPIs, given the interactions structure. It leverages supervised and semi-supervised feature-based classifiers, including our new Random Forest self-learning protocol. The classifiers are trained based on a dataset of comprehensive mutagenesis studies for 151 PPI complexes, with experimentally determined binding affinities of the mutant and wild-type interactions. Three classification problems were considered: (1) a 2-class problem (strengthening/weakening PPI mutations), (2) another 2-class problem (mutations that disrupt/preserve a PPI), and (3) a 3-class classification (detrimental/neutral/beneficial mutation effects). In total, 11 different supervised and semi-supervised classifiers were trained and assessed resulting in a promising performance, with the weighted f-measure ranging from 0.87 for Problem 1 to 0.70 for the most challenging Problem 3. By integrating prediction results of the 2-class classifiers into the 3-class classifier, we further improved its performance for Problem 3. To demonstrate the utility of SNP-IN tool, it was applied to study the nsSNP-induced rewiring of two disease-centered networks. The accurate and balanced performance of SNP-IN tool makes it readily available to study the rewiring of large-scale protein-protein interaction networks, and can be useful for functional annotation of disease-associated SNPs. SNIP-IN tool is freely accessible as a web-server at http://korkinlab.org/snpintool/.

Developing a Case-Based Reasoning Knowledge Repository To Support a Learning Community--An Example from the Technology Integration Community

People have long relied on storytelling for communicating ideas, transferring knowledge, and consequently making decisions. This paper describes the design and development of a case-based reasoning (CBR) knowledge repository including its case library and the search engine to support the technology integration community. CBR assumes that community knowledge can be captured in the form of stories (cases) so that the community members learn and solve problems by applying the lessons preserved in the stories to the current situation. The process and results of formative evaluation following the concept of participatory design are reported to set the context for further research and development.

A fast protein structure retrieval system using image-based distance matrices and multidimensional index

Indexing protein structures has been shown to provide a scalable solution for structure-to-structure comparisons in large protein structure retrieval systems. To conduct similarity searches against 46,075 polypeptide chains in a database with real-time responses, two critical issues must be addressed, information extraction and suitable indexing. In this paper, we apply computer vision techniques to extract the predominant information encoded in each 2D distance matrix, generated from 3D coordinates of protein chains. Distance matrices are capable of representing specific protein structural topologies, and similar proteins will generate similar matrices. Once meaningful features are extracted from distance images, an advanced indexing structure, entropy balanced statistical (EBS) k-d tree, can be utilized to index the multidimensional data. With a limited amount of training data from domain experts, namely structural classification of a subset of available protein chains, we apply various techniques in the pattern recognition field to determine clusters of proteins in the multi-dimensional feature space. Our system is able to recall search results in a ranked order from the protein database in seconds, exhibiting a reasonably high degree of precision.

Long identical multispecies elements in plant and animal genomes

Ultraconserved elements (UCEs) are DNA sequences that are 100% identical (no base substitutions, insertions, or deletions) and located in syntenic positions in at least two genomes. Although hundreds of UCEs have been found in animal genomes, little is known about the incidence of ultraconservation in plant genomes. Using an alignment-free information-retrieval approach, we have comprehensively identified all long identical multispecies elements (LIMEs), which include both syntenic and nonsyntenic regions, of at least 100 identical base pairs shared by at least two genomes. Among six animal genomes, we found the previously known syntenic UCEs as well as previously undescribed nonsyntenic elements. In contrast, among six plant genomes, we only found nonsyntenic LIMEs. LIMEs can also be classified as either simple (repetitive) or complex (nonrepetitive), they may occur in multiple copies in a genome, and they are often spread across multiple chromosomes. Although complex LIMEs were found in both animal and plant genomes, they differed significantly in their composition and copy number. Further analyses of plant LIMEs revealed their functional diversity, encompassing elements found near rRNA and enzyme-coding genes, as well as those found in transposons and noncoding DNA. We conclude that despite the common presence of LIMEs in both animal and plant lineages, the evolutionary processes involved in the creation and maintenance of these elements differ in the two groups and are likely attributable to several mechanisms, including transfer of genetic material from organellar to nuclear genomes, de novo sequence manufacturing, and purifying selection.

Countering imbalanced datasets to improve adverse drug event predictive models in labor and delivery

BACKGROUND The IOM report, Preventing Medication Errors, emphasizes the overall lack of knowledge of the incidence of adverse drug events (ADE). Operating rooms, emergency departments and intensive care units are known to have a higher incidence of ADE. Labor and delivery (L&D) is an emergency care unit that could have an increased risk of ADE, where reported rates remain low and under-reporting is suspected. Risk factor identification with electronic pattern recognition techniques could improve ADE detection rates. OBJECTIVE The objective of the present study is to apply Synthetic Minority Over Sampling Technique (SMOTE) as an enhanced sampling method in a sparse dataset to generate prediction models to identify ADE in women admitted for labor and delivery based on patient risk factors and comorbidities. RESULTS By creating synthetic cases with the SMOTE algorithm and using a 10-fold cross-validation technique, we demonstrated improved performance of the Naïve Bayes and the decision tree algorithms. The true positive rate (TPR) of 0.32 in the raw dataset increased to 0.67 in the 800% over-sampled dataset. CONCLUSION Enhanced performance from classification algorithms can be attained with the use of synthetic minority class oversampling techniques in sparse clinical datasets. Predictive models created in this manner can be used to develop evidence based ADE monitoring systems.