Abstract
In this chapter, the authors propose a new method belonging to content medical-based image retrieval approaches and that uses a set of region-based shape descriptors. The search engine discussed in this work allows the classification of newly acquired medical images into some well known categories and also to get the images that are more similar to a query image. The final goal is to help the medical staff to annotate these images. To achieve this task, the authors propose a set of three descriptors that are based on: (1) Hu, (2) Zernike moments, and (3) Fourier transform-based signature, which are considered as region descriptors. The advantage of using this kind of global descriptor is that they are very fast, real time, and they do not need any segmentation step. The authors propose also a comparative study between these three approaches. The search engines are tested by using a database composed of 75 images that have different sizes, and that are classified into five classes. The results provided by the proposed retrieval approaches are given with high precision. The comparison between the three approaches is achieved using classification matrices and the recall/precision curves. The three proposed retrieval approaches produce accurate results in real time. This proves the advantage of using global shape features as a preliminary classification step in an automated aided diagnosis system.
Top1. Introduction
Content Based Image Retrieval (CBIR) has been one of the most important research areas in the field of computer vision over the last 20 years. The huge amount of available multimedia data and the growing accessibility of visual information's are driving the need for thematic and easy access methods for allowing retrieve multimedia content; in the place of using text based multimedia retrieval (Müller, Michoux, Bandon & Geissbuhler, 2004).
Medical images are widely used to make and support the clinical decisions in hospitals and health centers. In medical practices, medical images are very useful and are used to provide enough information about diseases state, in order to recommend an effective treatment. A large volume of medical images are produced and stored in hospitals. However, these images are not exploited again after they have been used to make a first diagnosis. The annotation of image collections for retrieval is an active research area (Yao, Zhang, Antani, Long, & Thoma, 2008). Indeed, there are many problems such as the subjectivity of the annotation which is depending on human labelling and also the problem of choosing relevant keywords even when working with restricted vocabulary which need to be addressed.
If they are well annotated, medical images databases available in every health centers can be considered as a knowledge base concerning diseases and diagnosis. This information could be used for any medical diagnosis process. However, a large majority of information systems used in hospitals like PACS systems (Picture Archiving and Communication Systems) provide a response to the queries using only simple textual and clinical attributes (Hood & Scott, 2006). In the other hand, DICOM (Digital Imaging and Communications in Medicine), is a standard for image communication allowing patient information storage within the actual image(s). Although this storage system still having some problems of standardization.
The methods issued from Content Based Image Retrieval can solve this annotation problem, and can be used to assist the medical diagnosis. These kinds of applications belong to the new emerging challenges of medical image retrieval problems named Content Medical Based Image Retrieval (CMBIR). In this context, medical image databases can provide the possibility to search by image contents, and this by using the advantages of visual features. The CBIR techniques, which are used for image search and recognition, can give a solution that enriches and extends the functionalities and the efficiency of existing systems. We can notice that some important research in this field has evolved Content Based Image Retrieval (CBIR) into different scenarios such as arts, films industry, 3D retrieval, medical image retrieval, among others (Veltkamp, Burkhardt, & Kriegel, 2001; Sebe, Lew, Zhou, Huang, & Bakker, 2003; Tangelder & Veltkamp, 2008).
Key Terms in this Chapter
Search Engine: A software program that searches specific documents in a database and returns a list of items where the keywords were found.
Picture Archiving and Communication System (PACS): A combination of hardware and software dedicated to the short and long term storage, retrieval, management, distribution and presentation of images. The biggest consumers of PACS are hospitals. PACS main purpose is to replace hard film copies with digital images that can be used and seen by several different medical professionals simultaneously.
Content Medical Based Images Retrieval (CMBIR): The goal of Content Medical Based Images Retrieval (CMBIR) systems is to apply CBIR techniques to medical image databases. CMBIR approaches aim to assist the physician and doctors by predicting the disease of a particular case. In general, visually similar images correspond to the same disease category. By consulting the result of a CMBIR system, the specialist can gain more confidence in his/her decision or even consider other possibilities.
Content Medical Based Images Retrieval (CMBIR): The goal of Content Medical Based Images Retrieval (CMBIR) systems is to apply CBIR techniques to medical image databases. CMBIR approaches aim to assist the physician and doctors by predicting the disease of a particular case. In general, visually similar images correspond to the same disease category. By consulting the result of a CMBIR system, the specialist can gain more confidence in his/her decision or even consider other possibilities.
Computer-Aided Diagnosis (CAD): The purpose of Computer-Aided Diagnosis (CAD) procedures in medicine is to help and assist doctors in their interpretation of medical images. These procedures use the techniques issued from computer vision approaches. Automatic segmentation, quantification and also medical images retrieval are widely used in CAD.
Picture Archiving and Communication System (PACS): A combination of hardware and software dedicated to the short and long term storage, retrieval, management, distribution and presentation of images. The biggest consumers of PACS are hospitals. PACS main purpose is to replace hard film copies with digital images that can be used and seen by several different medical professionals simultaneously.
Search Engine: A software program that searches specific documents in a database and returns a list of items where the keywords were found.
Content-Based Image Retrieval (CBIR): Also known as Query By Image Content (QBIC), presents the technologies allowing to organize digital pictures by their visual features. They are based on the application of computer vision techniques to the image retrieval problem in large databases. Content-Based Image Retrieval (CBIR) consists of retrieving the most visually similar images to a given query image from a database of images.
Computer-Aided Diagnosis (CAD): The purpose of Computer-Aided Diagnosis (CAD) procedures in medicine is to help and assist doctors in their interpretation of medical images. These procedures use the techniques issued from computer vision approaches. Automatic segmentation, quantification and also medical images retrieval are widely used in CAD.
Digital Imaging and Communications in Medicine (DICOM): A global Information Technology standard that is used in hospitals. It is designed to ensure the interoperability of systems used for storing, printing, and transmitting information in medical imaging. It includes a file format definition and a network communication protocol. DICOM files can be exchanged between two entities that are capable of receiving image and patient data in DICOM format.
Descriptor: In an information retrieval system, a descriptor is a word or a characteristic feature used to identify an item (as a subject or document). In computer vision, visual descriptors present the descriptions of the visual features of the images or video contents. They describe, under invariance conditions, the elementary characteristics such as shape, color or texture.
Content-Based Image Retrieval (CBIR): Also known as Query By Image Content (QBIC), presents the technologies allowing to organize digital pictures by their visual features. They are based on the application of computer vision techniques to the image retrieval problem in large databases. Content-Based Image Retrieval (CBIR) consists of retrieving the most visually similar images to a given query image from a database of images.
Descriptor: In an information retrieval system, a descriptor is a word or a characteristic feature used to identify an item (as a subject or document). In computer vision, visual descriptors present the descriptions of the visual features of the images or video contents. They describe, under invariance conditions, the elementary characteristics such as shape, color or texture.
Digital Imaging and Communications in Medicine (DICOM): A global Information Technology standard that is used in hospitals. It is designed to ensure the interoperability of systems used for storing, printing, and transmitting information in medical imaging. It includes a file format definition and a network communication protocol. DICOM files can be exchanged between two entities that are capable of receiving image and patient data in DICOM format.