Image analysis and modeling in medical image computing. Recent developments and advances.

BACKGROUND: Medical image computing is of growing importance in medical diagnostics and image-guided therapy. Nowadays, image analysis systems integrating advanced image computing methods are used in practice e.g. to extract quantitative image parameters or to support the surgeon during a navigated intervention. However, the grade of automation, accuracy, reproducibility and robustness of medical image computing methods has to be increased to meet the requirements in clinical routine. OBJECTIVES: In the focus theme, recent developments and advances in the field of modeling and model-based image analysis are described. The introduction of models in the image analysis process enables improvements of image analysis algorithms in terms of automation, accuracy, reproducibility and robustness. Furthermore, model-based image computing techniques open up new perspectives for prediction of organ changes and risk analysis of patients. METHODS: Selected contributions are assembled to present latest advances in the field. The authors were invited to present their recent work and results based on their outstanding contributions to the Conference on Medical Image Computing BVM 2011 held at the University of Lübeck, Germany. All manuscripts had to pass a comprehensive peer review. RESULTS: Modeling approaches and model-based image analysis methods showing new trends and perspectives in model-based medical image computing are described. Complex models are used in different medical applications and medical images like radiographic images, dual-energy CT images, MR images, diffusion tensor images as well as microscopic images are analyzed. The applications emphasize the high potential and the wide application range of these methods. CONCLUSIONS: The use of model-based image analysis methods can improve segmentation quality as well as the accuracy and reproducibility of quantitative image analysis. Furthermore, image-based models enable new insights and can lead to a deeper understanding of complex dynamic mechanisms in the human body. Hence, model-based image computing methods are important tools to improve medical diagnostics and patient treatment in future.

Advances in medical image computing.

OBJECTIVES: Medical image computing has become a key technology in high-tech applications in medicine and an ubiquitous part of modern imaging systems and the related processes of clinical diagnosis and intervention. Over the past years significant progress has been made in the field, both on methodological and on application level. Despite this progress there are still big challenges to meet in order to establish image processing routinely in health care. In this issue, selected contributions of the German Conference on Medical Image Processing (BVM) are assembled to present latest advances in the field of medical image computing. METHODS: The winners of scientific awards of the German Conference on Medical Image Processing (BVM) 2008 were invited to submit a manuscript on their latest developments and results for possible publication in Methods of Information in Medicine. Finally, seven excellent papers were selected to describe important aspects of recent advances in the field of medical image processing. RESULTS: The selected papers give an impression of the breadth and heterogeneity of new developments. New methods for improved image segmentation, non-linear image registration and modeling of organs are presented together with applications of image analysis methods in different medical disciplines. Furthermore, state-of-the-art tools and techniques to support the development and evaluation of medical image processing systems in practice are described. CONCLUSIONS: The selected articles describe different aspects of the intense development in medical image computing. The image processing methods presented enable new insights into the patient's image data and have the future potential to improve medical diagnostics and patient treatment.

Integration of decentralized clinical data in a data warehouse: a service-oriented design and realization.

OBJECTIVES: In this paper we present a general concept and describe the difficulties for the integration of data from various clinical partners in one data warehouse using the Open European Nephrology Science Center (OpEN.SC) as an example. This includes a requirements analysis of the data integration process and also the design according to these requirements. METHODS: This conceptual approach based on the Rational Unified Process (RUP) and paradigm of Service-Oriented Architecture (SOA). RESULTS: Because we have to enhance the confidence of our partners in the OpEN.SC system and with this the willingness of them to participate, important requirements are controllability, transparency and security for all partners. Reusable and fine-grained components were found to be necessary when working with diverse data sources. With SOA the requested reusability is implemented easily. CONCLUSIONS: A key step in the development of a data integration process within such a health information system like OpEN.SC is to analyze the requirements. And to show that this is not only a theoretical work, we present a design - developed with RUP and SOA - which fulfills these requirements.

Advances in biomedical image analysis--past, present and future challenges.

Starting from raw data files coding eight bits of gray values per image pixel and identified with no more than eight characters to refer to the patient, the study, and technical parameters of the imaging modality, biomedical imaging has undergone manifold and rapid developments. Today, rather complex protocols such as Digital Imaging and Communications in Medicine (DICOM) are used to handle medical images. Most restrictions to image formation, visualization, storage and transfer have basically been solved and image interpretation now sets the focus of research. Currently, a method-driven modeling approach dominates the field of biomedical image processing, as algorithms for registration, segmentation, classification and measurements are developed on a methodological level. However, a further metamorphosis of paradigms has already started. The future of medical image processing is seen in task-oriented solutions integrated into diagnosis, intervention planning, therapy and follow-up studies. This alteration of paradigms is also reflected in the literature. As German activities are strongly tied to the international research, this change of paradigm is demonstrated by selected papers from the German annual workshop on medical image processing collected in this special issue.

Simulation and analysis of magnetic resonance elastography wave images using coupled harmonic oscillators and Gaussian local frequency estimation.

New methods for simulating and analyzing Magnetic Resonance Elastography (MRE) images are introduced. To simulate a two-dimensional shear wave pattern, the wave equation is solved for a field of coupled harmonic oscillators with spatially varying coupling and damping coefficients in the presence of an external force. The spatial distribution of the coupling and the damping constants are derived from an MR image of the investigated object. To validate the simulation as well as to derive the elasticity modules from experimental MRE images, the wave patterns are analyzed using a Local Frequency Estimation (LFE) algorithm based on Gauss filter functions with variable bandwidths. The algorithms are tested using an Agar gel phantom with spatially varying elasticity constants. Simulated wave patterns and LFE results show a high agreement with experimental data. Furthermore, brain images with estimated elasticities for gray and white matter as well as for exemplary tumor tissue are used to simulate experimental MRE data. The calculations show that already small distributions of pathologically changed brain tissue should be detectable by MRE even within the limit of relatively low shear wave excitation frequency around 0.2 kHz.

Volume-selective proton MR spectroscopy for in-vitro quantification of anticonvulsants.

Administration of anticonvulsant drugs is clinically monitored by checking seizure frequency and by determining the serum concentration of the drug. In a few reports, drug concentrations in brain parenchyma have been determined using ex vivo techniques. Little is known about the in vivo concentration in the brain parenchyma. Our goals were to characterise the NMR spectra of the anticonvulsants at therapeutic concentrations, to determine the minimum detectable concentrations, and to quantify the drugs noninvasively. Volume-selective 1H-MR spectroscopy (MRS) was performed under standard clinical conditions using a single-voxel STEAM (stimulated-echo acquisition mode) sequence at 1.5 T. Spectra of the anticonvulsants carbamazepine, phenobarbital, phenytoin and valproate were acquired in vitro in hydrous solutions at increasing dilution. Phenytoin, phenobarbital and valproate were detectable below maximum therapeutic serum concentrations. Within therapeutic ranges, there was good agreement between concentrations determined by 1H-MRS and those by standard fluorescence polarisation immunoassay. Due to the absence of signals of brain metabolites, the aromatic protons of phenobarbital, phenytoin and carbamazepine, with resonance lines around 7.4 ppm, allow the drugs to be detected. Valproate, with two resonances around 1.2 ppm, should be differentiable from potential brain metabolites using nonlinear analysis of the brain spectrum. Volume-selective 1H-MRS is therefore expected to be able to monitor anticonvulsant therapy in vivo.

Implementation of a dynamic platform-independent DICOM-server.

Hospital-wide image and patient data transfer within heterogeneous hard- and software infrastructures can be facilitated by using standardized communication protocols and data formats such as digital imaging and communications in medicine (DICOM). Each DICOM application entity (AE) usually provides a static and fixed set of services according to its functionality. However, certain security concepts, changing demands of medical users, or restricted hardware capabilities may be more easily addressed by applications that dynamically provide variable subsets of DICOM services. In a new approach, an object-oriented DICOM server framework was developed that served as a basis for assembling various DICOM applications. These applications may be set up dynamically to offer variable subsets of services at runtime, similar to "plug-ins". The framework was designed and implemented in Java in order to provide low-cost platform-independent solutions. As an example, a DICOM server was implemented and tested in a clinical application providing access to MR and CT images through a Java/DICOM viewer. Data retrieval was optimized by storing parts of the image acquisition and patient data into a relational database.

Histogram-based characterization of healthy and ischemic brain tissues using multiparametric MR imaging including apparent diffusion coefficient maps and relaxometry.

Decreased, renormalized, or increased values of the calculated apparent diffusion coefficient (ADC) are observed in stroke models. A quantitative description of corresponding tissue states using ADC values may be extended to include true relaxation times. A histogram-based segmentation is well suited for characterizing tissues according to specific parameter combinations irrespective of the heterogeneity found for human healthy and ischemic brain tissues. In a new approach, navigated diffusion-weighted images and ADC maps were incorporated into voxel-based parameter sets of relaxation times (T1, T2), and T1- or T2-weighted images, followed by a supervised histogram-based analysis. Healthy tissues were segmented by incorporating T1 relaxation into the data set, ischemic regions by combining T2- or diffusion-weighted images with ADC maps. Mean values of healthy and pathologic tissues were determined, spatial distributions of the parameter vectors were visualized using color-encoded overlays. One to six days after stroke, ischemic regions exhibited reduced relative mean ADC values.

Realization of security concepts for DICOM-based distributed medical services.

Exploiting distributed hard- and software resources for telemedicine requires a fast, secure, and platform-independent data exchange. Standards without inherent security mechanisms such as DICOM may ease non-authorized data access. Therefore, exemplary telemedical data streams were analyzed within the Berlin metropolitan area network using specialized magnetic resonance imaging techniques and distributed resources for data postprocessing. For secure DICOM communication both the Secure Socket Layer Protocol and a DICOM-conform partial encryption of patient-relevant data were implemented. Partial encryption exhibited the highest transfer rate and enabled a secure long-term storage. Different data streams between secured and unsecured networks were realized using partial encryption.

Prototype of a JAVA/DICOM image server with integrated findings and data security.

The transfer of medical data within heterogeneous hard- and software infrastructures requires platform-independent standardized protocols and data formats such as DICOM. To avoid costly vendor-specific solutions a DICOM server was implemented in JAVA thereby enabling the data access via internet browser technology. The most important patient and image acquisition information were extracted from the DICOM images and stored into a relational database. For an integrated view patient information such as radiological findings were transferred from the Radiological Information System (RIS) into the data base. Image data were accessed either by a fast preview tool or using a DICOM viewer. Since DICOM does not include inherent data security mechanisms, a second tool allowed the DICOM-conform encryption of DCIOM data for a secure long term storage on CD-R or across unsecure networks.

Virtual 3D cutting for bone segment extraction in maxillofacial surgery planning.

An important step toward our main goal of a completely computer-based maxillofacial surgical planning system is the availability of tools for the surgeon to define bone segments from skull and jaw bones. We have developed an easy-to-handle user interface that employs visual and force-feedback devices to define subvolumes of a patient's volume dataset. This interface is a main component of our maxillofacial surgical planning tool MeVisTo-Jaw [1]. The defined subvolumes together with their spatial arrangements lead to an operation plan.

Units of measure in clinical information systems.

The authors surveyed existing standard codes for units of measures, such as ISO 2955, ANSI X3.50, and Health Level 7's ISO+. Because these standards specify only the character representation of units, the authors developed a semantic model for units based on dimensional analysis. Through this model, conversion between units and calculations with dimensioned quantities become as simple as calculating with numbers. All atomic symbols for prefixes and units are defined in one small table. Huge permutated conversion tables are not required. This method is also simple enough to be widely implementable in today's information systems. To promote the application of the method the authors provide an open-source implementation of this method in JAVA. All existing code standards for units, however, are incomplete for practical use and require substantial changes to correct their many ambiguities. The authors therefore developed a code for units that is much more complete and free from ambiguities.

Implementing HL7: from the standard's specification to production application.

A C++ implementation of the HL7 health-care data interchange standard was developed by automatic methods applied to the authoritative specification of the standard. The reusable class library thus created presents an intuitive, flexible, and easy-to-use application programming interface to the HL7 protocol. This allows HL7 applications to be developed quickly while a high conformance to the standard is ensured.

Knowledge-based system ADNEXPERT to assist the sonographic diagnosis of adnexal tumors.

ADNEXPERT is a knowledge-based system for the computer-assisted ultrasound diagnosis of adnexal tumors. In a case-based approach, ADNEXPERT used histopathologic and sonographic data from 2,290 adnexal tumors. After an ultrasound examination, the gynecologist interacts with the system. A maximum of 15 questions are posed; all but one question (age) relate to the sonographic findings. The help system gives online access to an ultrasound image library. Once the dialogue is complete, ADNEXPERT assesses the adnexal tumor pathology and makes a histological classification. A certainty factor (CF) model is used for knowledge representation. The CFs of the knowledge base are computed from the case database. During system evaluation, the accuracy of ADNEXPERT was tested by 69 new adnexal tumor cases, for which verified histopathological diagnoses were available. ADNEXPERT accurately assessed pathology in 49 cases (71%); in 10 cases (14%) correct indications to pathology were given; no diagnostic hints were attained in 2 cases (3%); and 8 cases (12%) were falsely diagnosed. Based on the positive results of the evaluation, ADNEXPERT will be tested under clinical conditions.

[RADIOLIS. A radiologic instruction and training system for systematic evaluation of roentgen images exemplified by focal bone lesions]. / RADIOLIS. Ein radiologisches Instruktions- und Trainingssystem zur systematischen Befundung von Röntgenbildern am Beispiel fokaler Knochenläsionen.

RADIOLIS is an instructional software system for radiological training that guides the user through a systematic analysis of an image. The backbone of the analysis is a detailed and structured questionnaire which, as applied to our example, includes all relevant differential-diagnostic criteria for focal bone lesions. The variation range of the radiomorphologic characteristics in the questionnaire is introduced to the user with digitized X-ray images, making the simulation of an image analysis possible. In preparation, diverse teaching/learning methods reinforced by instructional psychology were incorporated during the system's development. At the core of the training system are an expandable image database and a database containing descriptions of image material formulated by radiology experts. The field of focal bone lesions, for instance, has already been integrated into RADIOLIS. A graphic user interface and an interactive dialogue enable both comfortable and simple use of the system and the user-specific structuring of the learning process.

Differentiation of normal and pathologic brain structures in MRI using exact T1 and T2 values followed by a multidimensional cluster analysis.

A new method to differentiate and classify normal and pathologic brain tissue in Magnetic Resonance Imaging (MRI) is introduced. With a special pulse sequence, exact T1 and T2 values are simultaneously acquired, where the T2 values may show a biexponential decay behavior. The automatic segmentation of the tissue is based on a histogram analysis of the multidimensional parameters. The tissue-characterizing parameters are stored in a database, which serves as an objective classification of different tissues. A pilot study with 15 healthy volunteers and 100 patients showed a good differentiation between normal tissue like white and gray matter, cerebrospinal fluid, muscle, and fat, thereby revealing a biexponential decay behavior of fat. Pathologic tissue-like edema and meningioma could be classified with high accuracy, while glioblastoma or astrocytoma were difficult to classify because of their non-homogeneous structures.

Texture-based X-ray image segmentation by topological map.

The texture-based segmentation of x-ray images of focal bone lesions using topological maps is introduced. Texture characteristics are described by image-point correlation of feature images to feature vectors. For the segmentation, the topological map is labeled using an improved labeling strategy. Results of the technique are demonstrated on original and synthetic x-ray images and quantified with the aid of quality measures. In addition, a classifier-specific contribution analysis is applied for assessing the feature area.

[New developments in the computer-assisted diagnosis of focal bone lesions]. / Neue Ansätze zur computerassistierten Diagnose von fokalen Knochenläsionen.

Since the 1960s there have been numerous reports in the literature about computer-aided medical diagnosis and therapy. However, so far computer-aided diagnosis has not become important in the clinical routine or in the field of radiology. This paper describes the hitherto existing proposals for expert systems to evaluate focal bone lesions and presents a new concept. The basic ideas of this program are: (1) a PC-based interactive dialogue along on the basis of a defined questionnaire; (2) the use of certainty factors to handle uncertainty and vagueness in evaluating the different radiological findings and their correlation with diagnoses; (3) the link between visual information and the questionnaire. The aims of such expert systems are (a) a complete and systematic analysis of the films, (b) independency of association with previously diagnosed cases in reporting the films, (c) the procurement and use of radiological descriptive terms, and (d) aid in the radiological evaluation of focal bone lesions.

In vivo NMR T2 relaxation of experimental brain tumors in the cat: a multiparameter tissue characterization.

Experimental gliomas (F98) were inoculated in cat brain for the systematic study of their in vivo T2 relaxation time behavior. With a CPMG multi-echo imaging sequence, a train of 16 echoes was evaluated to obtain the transverse relaxation time and the magnetization M(0) at time t = 0. The magnetization decay curves were analyzed for biexponentiality. All tissues showed monoexponential T2, only that of the ventricular fluid and part of the vital tumor tissue were biexponential. Based on these NMR relaxation parameters the tissues were characterized, their correct assignment being assured by comparison with histological slices. T2 of normal grey and white matter was 74 +/- 6 and 72 +/- 6 msec, respectively. These two tissue types were distinguished through M(0) which for white matter was only 0.88 of the intensity of grey matter in full agreement with water content, determined from tissue specimens. At the time of maximal tumor growth and edema spread a tissue differentiation was possible in NMR relaxation parameter images. Separation of the three tissue groups of normal tissue, tumor and edema was based on T2 with T2(normal) < T2(tumor) < T2(edema). Using M(0) as a second parameter the differentiation was supported, in particular between white matter and tumor or edema. Animals were studied at 1-4 wk after tumor implantation to study tumor development. The magnetization M(0) of both tumor and peritumoral edema went through a maximum between the second and third week of tumor growth. T2 of edema was maximal at the same time with 133 +/- 4 msec, while the relaxation time of tumor continued to increase during the whole growth period, reaching values of 114 +/- 12 msec at the fourth week. Thus, a complete characterization of pathological tissues with NMR relaxometry must include a detailed study of the developmental changes of these tissues to assure correct experimental conditions for the goal of optimal contrast between normal and pathological regions in the NMR images.

[Possible applications of a method for 3D reconstruction of CT image data. OP-planning, finishing of individual alloplastic implants for replacement of cranial and maxillofacial bone structures]. / Einsatzmöglichkeiten der 3D-Rekonstruktion von CT-Daten. OP-Planung, Fertigung individueller alloplastischer Implantate zum Ersatz kranialer und maxillofazialer knöcherner Strukturen.

A method for the three-dimensional reconstruction of CT-image data permits manufacturing of natural size anatomical models thus enabling preoperative measurements of relevant parameters (length, areas, and volumes). The methods described allow for new perspectives in preoperative planning of dental implants, particularly for the maxillary problem area. The use of 3D image data for manufacturing individual alloplastic implants simplifies the exact reconstruction of symmetrical bone structures, which up to now was impossible to achieve with conventional methods. Applications in the fields of orthodontic surgery and in neurological surgery are suggested. Refinement of available techniques is possible by means of additional segmentation of soft tissues, and by implementation of more sophisticated software for better visualisation and smoothly shaped surfaces. Future applications may even include aesthetic surgery.