Use of an Online Photo Management and Sharing Application to Create a Searchable Digital Dental Radiograph Repository

Problem: Dental radiographs generally display one or more findings/diagnoses, and are linked to a unique set of patient demographics, medical history and other findings not represented by the image. However, this information is not associated with radiographs in any type of meta format, and images are not searchable based on any clinical criteria (1,2). The purpose of this pilot study is to create an online, searchable data repository of dental radiographs to be used for patient care, teaching and research. [See PDF for complete abstract]

A Novel Approach for Multispectral Satellite Image Classification Based on the Bat Algorithm

Among the multiple advantages and applications of remote sensing, one of the most important uses is to solve the problem of crop classification, i.e., differentiating between various crop types. Satellite images are a reliable source for investigating the temporal changes in crop cultivated areas. In this letter, we propose a novel bat algorithm (BA)-based clustering approach for solving crop type classification problems using a multispectral satellite image. The proposed partitional clustering algorithm is used to extract information in the form of optimal cluster centers from training samples. The extracted cluster centers are then validated on test samples. A real-time multispectral satellite image and one benchmark data set from the University of California, Irvine (UCI) repository are used to demonstrate the robustness of the proposed algorithm. The performance of the BA is compared with two other nature-inspired metaheuristic techniques, namely, genetic algorithm and particle swarm optimization. The performance is also compared with the existing hybrid approach such as the BA with K-means. From the results obtained, it can be concluded that the BA can be successfully applied to solve crop type classification problems.

Online document repository for GAH Website (2016)

Resources created at University of Southampton for the module GIS for the Analysis of Health (GAH)

Estimating Ribeye Area Using a Longitudinal Ultrasound Image of the 12th and 13th Rib Section

A study was performed to determine the feasability of using a measurement of ribeye depth (RED) from a longitudinal ultrasound image to estimate ribeye area (REA). The correlation between RED obtained with ultrasound and REA from a tracing was high for both implanted (r = .49) and non-implanted (r = .45) steers. The mean bias between predicted REA and actual REA was not different from zero. This analysis shows that RED could be an accurate indicator of REA.

The Virtual Skeleton Database: An Open Access Repository for Biomedical Research and Collaboration

Background: Statistical shape models are widely used in biomedical research. They are routinely implemented for automatic image segmentation or object identification in medical images. In these fields, however, the acquisition of the large training datasets, required to develop these models, is usually a time-consuming process. Even after this effort, the collections of datasets are often lost or mishandled resulting in replication of work. Objective: To solve these problems, the Virtual Skeleton Database (VSD) is proposed as a centralized storage system where the data necessary to build statistical shape models can be stored and shared. Methods: The VSD provides an online repository system tailored to the needs of the medical research community. The processing of the most common image file types, a statistical shape model framework, and an ontology-based search provide the generic tools to store, exchange, and retrieve digital medical datasets. The hosted data are accessible to the community, and collaborative research catalyzes their productivity. Results: To illustrate the need for an online repository for medical research, three exemplary projects of the VSD are presented: (1) an international collaboration to achieve improvement in cochlear surgery and implant optimization, (2) a population-based analysis of femoral fracture risk between genders, and (3) an online application developed for the evaluation and comparison of the segmentation of brain tumors. Conclusions: The VSD is a novel system for scientific collaboration for the medical image community with a data-centric concept and semantically driven search option for anatomical structures. The repository has been proven to be a useful tool for collaborative model building, as a resource for biomechanical population studies, or to enhance segmentation algorithms.

The life sciences Global Image Database (GID)

Although a vast amount of life sciences data is generated in the form of images, most scientists still store images on extremely diverse and often incompatible storage media, without any type of metadata structure, and thus with no standard facility with which to conduct searches or analyses. Here we present a solution to unlock the value of scientific images. The Global Image Database (GID) is a web-based (http://www.g wer.ch/qv/gid/gid.htm) structured central repository for scientific annotated images. The GID was designed to manage images from a wide spectrum of imaging domains ranging from microscopy to automated screening. The annotations in the GID define the source experiment of the images by describing who the authors of the experiment are, when the images were created, the biological origin of the experimental sample and how the sample was processed for visualization. A collection of experimental imaging protocols provides details of the sample preparation, and labeling, or visualization procedures. In addition, the entries in the GID reference these imaging protocols with the probe sequences or antibody names used in labeling experiments. The GID annotations are searchable by field or globally. The query results are first shown as image thumbnail previews, enabling quick browsing prior to original-sized annotated image retrieval. The development of the GID continues, aiming at facilitating the management and exchange of image data in the scientific community, and at creating new query tools for mining image data.

Design and implementation of a multisite data repository for pediatric epilepsy

This dissertation established a software-hardware integrated design for a multisite data repository in pediatric epilepsy. A total of 16 institutions formed a consortium for this web-based application. This innovative fully operational web application allows users to upload and retrieve information through a unique human-computer graphical interface that is remotely accessible to all users of the consortium. A solution based on a Linux platform with My-SQL and Personal Home Page scripts (PHP) has been selected. Research was conducted to evaluate mechanisms to electronically transfer diverse datasets from different hospitals and collect the clinical data in concert with their related functional magnetic resonance imaging (fMRI). What was unique in the approach considered is that all pertinent clinical information about patients is synthesized with input from clinical experts into 4 different forms, which were: Clinical, fMRI scoring, Image information, and Neuropsychological data entry forms. A first contribution of this dissertation was in proposing an integrated processing platform that was site and scanner independent in order to uniformly process the varied fMRI datasets and to generate comparative brain activation patterns. The data collection from the consortium complied with the IRB requirements and provides all the safeguards for security and confidentiality requirements. An 1-MR1-based software library was used to perform data processing and statistical analysis to obtain the brain activation maps. Lateralization Index (LI) of healthy control (HC) subjects in contrast to localization-related epilepsy (LRE) subjects were evaluated. Over 110 activation maps were generated, and their respective LIs were computed yielding the following groups: (a) strong right lateralization: (HC=0%, LRE=18%), (b) right lateralization: (HC=2%, LRE=10%), (c) bilateral: (HC=20%, LRE=15%), (d) left lateralization: (HC=42%, LRE=26%), e) strong left lateralization: (HC=36%, LRE=31%). Moreover, nonlinear-multidimensional decision functions were used to seek an optimal separation between typical and atypical brain activations on the basis of the demographics as well as the extent and intensity of these brain activations. The intent was not to seek the highest output measures given the inherent overlap of the data, but rather to assess which of the many dimensions were critical in the overall assessment of typical and atypical language activations with the freedom to select any number of dimensions and impose any degree of complexity in the nonlinearity of the decision space.

Image-Guided Precision Manipulation of Cells and Nanoparticles in Microfluidics

Manipulation of single cells and particles is important to biology and nanotechnology. Our electrokinetic (EK) tweezers manipulate objects in simple microfluidic devices using gentle fluid and electric forces under vision-based feedback control. In this dissertation, I detail a user-friendly implementation of EK tweezers that allows users to select, position, and assemble cells and nanoparticles. This EK system was used to measure attachment forces between living breast cancer cells, trap single quantum dots with 45 nm accuracy, build nanophotonic circuits, and scan optical properties of nanowires. With a novel multi-layer microfluidic device, EK was also used to guide single microspheres along complex 3D trajectories. The schemes, software, and methods developed here can be used in many settings to precisely manipulate most visible objects, assemble objects into useful structures, and improve the function of lab-on-a-chip microfluidic systems.

Leveraging Multiple Features for Image Retrieval and Matching

The goal of image retrieval and matching is to find and locate object instances in images from a large-scale image database. While visual features are abundant, how to combine them to improve performance by individual features remains a challenging task. In this work, we focus on leveraging multiple features for accurate and efficient image retrieval and matching. We first propose two graph-based approaches to rerank initially retrieved images for generic image retrieval. In the graph, vertices are images while edges are similarities between image pairs. Our first approach employs a mixture Markov model based on a random walk model on multiple graphs to fuse graphs. We introduce a probabilistic model to compute the importance of each feature for graph fusion under a naive Bayesian formulation, which requires statistics of similarities from a manually labeled dataset containing irrelevant images. To reduce human labeling, we further propose a fully unsupervised reranking algorithm based on a submodular objective function that can be efficiently optimized by greedy algorithm. By maximizing an information gain term over the graph, our submodular function favors a subset of database images that are similar to query images and resemble each other. The function also exploits the rank relationships of images from multiple ranked lists obtained by different features. We then study a more well-defined application, person re-identification, where the database contains labeled images of human bodies captured by multiple cameras. Re-identifications from multiple cameras are regarded as related tasks to exploit shared information. We apply a novel multi-task learning algorithm using both low level features and attributes. A low rank attribute embedding is joint learned within the multi-task learning formulation to embed original binary attributes to a continuous attribute space, where incorrect and incomplete attributes are rectified and recovered. To locate objects in images, we design an object detector based on object proposals and deep convolutional neural networks (CNN) in view of the emergence of deep networks. We improve a Fast RCNN framework and investigate two new strategies to detect objects accurately and efficiently: scale-dependent pooling (SDP) and cascaded rejection classifiers (CRC). The SDP improves detection accuracy by exploiting appropriate convolutional features depending on the scale of input object proposals. The CRC effectively utilizes convolutional features and greatly eliminates negative proposals in a cascaded manner, while maintaining a high recall for true objects. The two strategies together improve the detection accuracy and reduce the computational cost.

Learning Binary Code Representations for Effective and Efficient Image Retrieval

The size of online image datasets is constantly increasing. Considering an image dataset with millions of images, image retrieval becomes a seemingly intractable problem for exhaustive similarity search algorithms. Hashing methods, which encodes high-dimensional descriptors into compact binary strings, have become very popular because of their high efficiency in search and storage capacity. In the first part, we propose a multimodal retrieval method based on latent feature models. The procedure consists of a nonparametric Bayesian framework for learning underlying semantically meaningful abstract features in a multimodal dataset, a probabilistic retrieval model that allows cross-modal queries and an extension model for relevance feedback. In the second part, we focus on supervised hashing with kernels. We describe a flexible hashing procedure that treats binary codes and pairwise semantic similarity as latent and observed variables, respectively, in a probabilistic model based on Gaussian processes for binary classification. We present a scalable inference algorithm with the sparse pseudo-input Gaussian process (SPGP) model and distributed computing. In the last part, we define an incremental hashing strategy for dynamic databases where new images are added to the databases frequently. The method is based on a two-stage classification framework using binary and multi-class SVMs. The proposed method also enforces balance in binary codes by an imbalance penalty to obtain higher quality binary codes. We learn hash functions by an efficient algorithm where the NP-hard problem of finding optimal binary codes is solved via cyclic coordinate descent and SVMs are trained in a parallelized incremental manner. For modifications like adding images from an unseen class, we propose an incremental procedure for effective and efficient updates to the previous hash functions. Experiments on three large-scale image datasets demonstrate that the incremental strategy is capable of efficiently updating hash functions to the same retrieval performance as hashing from scratch.

Efficient Image and Video Representations for Retrieval

Image (Video) retrieval is an interesting problem of retrieving images (videos) similar to the query. Images (Videos) are represented in an input (feature) space and similar images (videos) are obtained by finding nearest neighbors in the input representation space. Numerous input representations both in real valued and binary space have been proposed for conducting faster retrieval. In this thesis, we present techniques that obtain improved input representations for retrieval in both supervised and unsupervised settings for images and videos. Supervised retrieval is a well known problem of retrieving same class images of the query. We address the practical aspects of achieving faster retrieval with binary codes as input representations for the supervised setting in the first part, where binary codes are used as addresses into hash tables. In practice, using binary codes as addresses does not guarantee fast retrieval, as similar images are not mapped to the same binary code (address). We address this problem by presenting an efficient supervised hashing (binary encoding) method that aims to explicitly map all the images of the same class ideally to a unique binary code. We refer to the binary codes of the images as `Semantic Binary Codes' and the unique code for all same class images as `Class Binary Code'. We also propose a new class­ based Hamming metric that dramatically reduces the retrieval times for larger databases, where only hamming distance is computed to the class binary codes. We also propose a Deep semantic binary code model, by replacing the output layer of a popular convolutional Neural Network (AlexNet) with the class binary codes and show that the hashing functions learned in this way outperforms the state­ of ­the art, and at the same time provide fast retrieval times. In the second part, we also address the problem of supervised retrieval by taking into account the relationship between classes. For a given query image, we want to retrieve images that preserve the relative order i.e. we want to retrieve all same class images first and then, the related classes images before different class images. We learn such relationship aware binary codes by minimizing the similarity between inner product of the binary codes and the similarity between the classes. We calculate the similarity between classes using output embedding vectors, which are vector representations of classes. Our method deviates from the other supervised binary encoding schemes as it is the first to use output embeddings for learning hashing functions. We also introduce new performance metrics that take into account the related class retrieval results and show significant gains over the state­ of­ the art. High Dimensional descriptors like Fisher Vectors or Vector of Locally Aggregated Descriptors have shown to improve the performance of many computer vision applications including retrieval. In the third part, we will discuss an unsupervised technique for compressing high dimensional vectors into high dimensional binary codes, to reduce storage complexity. In this approach, we deviate from adopting traditional hyperplane hashing functions and instead learn hyperspherical hashing functions. The proposed method overcomes the computational challenges of directly applying the spherical hashing algorithm that is intractable for compressing high dimensional vectors. A practical hierarchical model that utilizes divide and conquer techniques using the Random Select and Adjust (RSA) procedure to compress such high dimensional vectors is presented. We show that our proposed high dimensional binary codes outperform the binary codes obtained using traditional hyperplane methods for higher compression ratios. In the last part of the thesis, we propose a retrieval based solution to the Zero shot event classification problem - a setting where no training videos are available for the event. To do this, we learn a generic set of concept detectors and represent both videos and query events in the concept space. We then compute similarity between the query event and the video in the concept space and videos similar to the query event are classified as the videos belonging to the event. We show that we significantly boost the performance using concept features from other modalities.