Definition of an automated Content-Based Image Retrieval (CBIR) system for the comparison of dermoscopic images of pigmented skin lesions

Department of Biochemistry, Section of Pathology, Second University of Naples, Naples, Italy.
BioMedical Engineering OnLine (Impact Factor: 1.75). 09/2009; 8:18. DOI: 10.1186/1475-925X-8-18
Source: PubMed

ABSTRACT New generations of image-based diagnostic machines are based on digital technologies for data acquisition; consequently, the diffusion of digital archiving systems for diagnostic exams preservation and cataloguing is rapidly increasing. To overcome the limits of current state of art text-based access methods, we have developed a novel content-based search engine for dermoscopic images to support clinical decision making.
To this end, we have enrolled, from 2004 to 2008, 3415 caucasian patients and collected 24804 dermoscopic images corresponding to 20491 pigmented lesions with known pathology. The images were acquired with a well defined dermoscopy system and stored to disk in 24-bit per pixel TIFF format using interactive software developed in C++, in order to create a digital archive.
The analysis system of the images consists in the extraction of the low-level representative features which permits the retrieval of similar images in terms of colour and texture from the archive, by using a hierarchical multi-scale computation of the Bhattacharyya distance of all the database images representation with respect to the representation of user submitted (query).
The system is able to locate, retrieve and display dermoscopic images similar in appearance to one that is given as a query, using a set of primitive features not related to any specific diagnostic method able to visually characterize the image. Similar search engine could find possible usage in all sectors of diagnostic imaging, or digital signals, which could be supported by the information available in medical archives.

  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: Content-based medical image retrieval (CBMIR) system enables medical practitioners to perform fastdiagnosis through quantitative assessment of the visual information of various modalities. In this paper, a more robust CBMIR system that deals with both cervical and lumbar vertebraeirregularity is afforded. It comprises three main phases, namely modelling, indexing and retrievalof the vertebrae image. The main tasks in the modelling phase are to improve, enhance thevisibility of the x-ray image for better segmentation results using active shape model (ASM). Thesegmented vertebral fractures are then characterized in the indexing phase using region-based fracturecharacterization (RB-FC) and contour-based fracture characterization (CB-FC). Upon a query, thecharacterized features are compared to the query image. Effectiveness of the retrieval phase isdetermined by its retrieval, thus, we propose an integration of the predictor model based crossvalidation neural network (PMCVNN) and similarity matching (SM) in this stage. The PMCVNN task is to identify the correct vertebral irregularity class through classification allowing the SM process tobe more efficient. Retrieval performance between the proposed and the standard retrieval architecturesare then compared using an retrieval precision (Pr@M) and average group score (AGS) measures. Experimental results show that the new integrated retrieval architecture performs better than those ofthe standard CBMIR architecture with retrieval results of cervical (AGS > 87%) and lumbar (AGS >82%) datasets. The proposed CBMIR architecture shows encouraging results with high Pr@M accuracy. As a result,images from the same visualization class are returned for further used by the medical personnel.
    BioMedical Engineering OnLine 01/2015; 14(1):6. DOI:10.1186/1475-925X-14-6 · 1.75 Impact Factor
  • [Show abstract] [Hide abstract]
    ABSTRACT: Web information retrieval has become one of the most important applications of the Web. Indeed, due to the abundance and velocity with which data has been and is put on the Web, retrieving the information appropriate for a particular purpose has become both challenging and an almost daily performed task. The applications where information found on the Web is between important and crucial, is needed, is critical, or is used as a basis for, say, decision making are numerous and manifold, and they keep growing. As the editors of this book state in the preface: “While information retrieval was developed within the librarians’ community well before the use of computers, its importance boosted at the turn of the century, with the diffusion of the World Wide Web. Big players in the computer industry, such as Google and Yahoo!, were the primary contributors of technology for fast access to Web information. Searching capabilities are now integrated in most information systems, ranging from business management software and customer relationship systems to social networks and mobile phone applications.” So it comes as no surprise that an understanding of what happens in Web information retrieval from a technical perspective is crucial for anybody studying or working in Computer Science. The intention of this book is to survey the field from a variety of angles. Ceri and his co-authors take their readers from the foundations of modern information retrieval to advanced challenges of Web information retrieval. Their book is divided into three parts: The first part addresses the principles of information retrieval and provides a systematic and compact description of basic techniques (including binary, vector space and probabilistic models, classification and clustering, natural language processing for search). The second part addresses the foundational aspects of Web information retrieval and discusses the general architecture of search engines (with a focus on the crawling and indexing processes), describes link analysis methods (specifically Page Rank and HITS), addresses recommendation and diversification, and finally covers advertising in search. The third part describes advanced aspects of Web search, from several directions. Topics in this part include publishing data on the Web, meta-search and multi-domain search, semantic search, search in the context of multimedia data, as well as human computation and crowd search. With this selection of topic, the book covers not only a wide range, but everything that is essential to the topic of Web information retrieval. It is not a replacement for a classical textbook on information retrieval, but it takes this general theme into the various corners that a relevant to the Web these days. All chapters are relatively short and hence easy to digest, yet mastering the topic dealt with in a chapter will almost certainly require additional study. Nevertheless, this book is an excellent starting point into the field of Web information retrieval, and can be recommended for classroom use. Accompanying material for instructors is available from
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: Background It is often difficult to differentiate early melanomas from benign melanocytic nevi even by expert dermatologists, and the task is even more challenging for primary care physicians untrained in dermatology and dermoscopy. A computer system can provide an objective and quantitative evaluation of skin lesions, reducing subjectivity in the diagnosis. Objective Our objective is to make a low-cost computer aided diagnostic tool applicable in primary care based on a consumer grade camera with attached dermatoscope, and compare its performance to that of experienced dermatologists. Methods and Material We propose several new image-derived features computed from automatically segmented dermoscopic pictures. These are related to the asymmetry, color, border, geometry, and texture of skin lesions. The diagnostic accuracy of the system is compared with that of three dermatologists. Results With a data set of 206 skin lesions, 169 benign and 37 melanomas, the classifier was able to provide competitive sensitivity (86%) and specificity (52%) scores compared with the sensitivity (85%) and specificity (48%) of the most accurate dermatologist using only dermoscopic images. Conclusion We show that simple statistical classifiers can be trained to provide a recommendation on whether a pigmented skin lesion requires biopsy to exclude skin cancer with a performance that is comparable to and exceeds that of experienced dermatologists.
    Artificial Intelligence in Medicine 12/2013; DOI:10.1016/j.artmed.2013.11.006 · 1.36 Impact Factor

Full-text (3 Sources)

Available from
Jun 6, 2014