Non-invasive diagnosing malignant melanoma by multi-spectral optical Nevoscope.

Malignant melanoma is one of the most fatal forms of skin cancer. There are two significant signs indicating malignancy of a melanoma, that is, abnormal melanin distribution in dermis layer and a peripheral blood net. We developed a multi-spectral optical Nevoscope aimed to diagnosing malignant melanoma non-invasively. An algorithm is proposed to reconstruct the melanoma in terms of Nevoscope geometry. The algorithm has been verified on an optical tumor model at 580 nm and 800 nm. The reconstructed melanoma is consistent with the tumor model which suggests a great potential of using Nevoscope to investigate malignant melanoma.

Planar visual fusion scintigraphy.

Planar scintigraphy, while providing useful information about the distribution of a particular radiopharmaceutical being imaged, often does not provide adequate information about the surrounding anatomical structures, thereby complicating diagnosis. We have therefore explored a means of fusing planar scintigraphic images with visual photographic images to supply an anatomic correlate to regions of radiopharmaceutical accumulation. The digital visual image will provide a context for the relevant structures in the scintigraphic image. Phantom data confirm registration accuracy to within 1 pixel. Inaccuracy of camera-patient distance results in <1% image size change per cm height error. Initial clinical imaging has subjectively been very useful in low background applications such as lymphoscintigraphy, whole body I-131 NaI imaging for thyroid cancer and In-111 white blood cell infection imaging.

Three dimensional reconstruction of coronary arteries from two views.

Geometric representation and measurements of localized lumen stenosis of coronary arteries are important considerations in the diagnosis of cardiovascular diseases. This discrete narrowing of the arteries typically impairs blood flow in regions of the heart, and can be present along the entire length of the artery. Three-dimensional (3-D) reconstruction of coronary arterial tree allows clinician to visualize vascular geometry. Three-dimensional representation of tree topology facilitates calculation of hemodynamic measurements to study myocardial infarction and stenosis. The 3-D arterial tree, computed from two views, can provide more information about the tree geometry than individual views. In this paper, a 3-step algorithm for 3-D reconstruction of arterial tree using two standard views is presented. The first step is a multi-resolution segmentation of the coronary vessels followed by medial-axis detection along the entire arterial tree for both views. In the second step, arterial trees from the two views are registered using medial-axis representation at the coarsest resolution level to obtain an initial 3-D reconstruction. This initial reconstruction at the coarsest level is then modified using 3-D geometrical a priori information. In the third step, the modified reconstruction is projected on the next higher-resolution segmented medial-axis representation and an updated reconstruction is obtained at the higher resolution. The process is iterated until the final 3-D reconstruction is obtained at the finest resolution level. Linear programming based constrained optimization method is used for registering two views at the coarse resolution. This is followed by a Tree-Search method for registering detailed branches at higher resolutions. The automated 3-D reconstruction method was evaluated on computer-simulated as well as human angiogram data. Results show that the automated 3-D reconstruction method provided good registration of computer-simulated data. On human angiogram data, the computed 3-D reconstruction matched well with manual registration.

Wavelet based multiresolution expectation maximization image reconstruction algorithm for positron emission tomography.

Maximum Likelihood (ML) estimation based Expectation Maximization (EM) [IEEE Trans Med Imag, MI-1 (2) (1982) 113] reconstruction algorithm has shown to provide good quality reconstruction for positron emission tomography (PET). Our previous work [IEEE Trans Med Imag, 7(4) (1988) 273; Proc IEEE EMBS Conf, 20(2/6) (1998) 759] introduced the multigrid (MG) and multiresolution (MR) concept for PET image reconstruction using EM. This work transforms the MGEM and MREM algorithm to a Wavelet based Multiresolution EM (WMREM) algorithm by extending the concept of switching resolutions in both image and data spaces. The MR data space is generated by performing a 2D-wavelet transform on the acquired tube data that is used to reconstruct images at different spatial resolutions. Wavelet transform is used for MR reconstruction as well as adapted in the criterion for switching resolution levels. The advantage of the wavelet transform is that it provides very good frequency and spatial (time) localization and allows the use of these coarse resolution data spaces in the EM estimation process. The MR algorithm recovers low-frequency components of the reconstructed image at coarser resolutions in fewer iterations, reducing the number of iterations required at finer resolution to recover high-frequency components. This paper also presents the design of customized biorthogonal wavelet filters using the lifting method that are used for data decomposition and image reconstruction and compares them to other commonly known wavelets.

Multi-level adaptive segmentation of multi-parameter MR brain images.

MR brain image segmentation into several tissue classes is of significant interest to visualize and quantify individual anatomical structures. Traditionally, the segmentation is performed manually in a clinical environment that is operator dependent and may be difficult to reproduce. Though several algorithms have been investigated in the literature for computerized automatic segmentation of MR brain images, they are usually targeted to classify image into a limited number of classes such as white matter, gray matter, cerebrospinal fluid and specific lesions. We present a novel model-based method for the automatic segmentation and classification of multi-parameter MR brain images into a larger number of tissue classes of interest. Our model employs 15 brain tissue classes instead of the commonly used set of four classes, which were of clinical interest to neuroradiologists for following-up with patients suffering from cerebrovascular deficiency (CVD) and/or stroke. The model approximates the spatial distribution of tissue classes by a Gauss Markov random field and uses the maximum likelihood method to estimate the class probabilities and transitional probabilities for each pixel of the image. Multi-parameter MR brain images with T(1), T(2), proton density, Gd+T(1), and perfusion imaging were used in segmentation and classification. In the development of the segmentation model, true class-membership of measured parameters was determined from manual segmentation of a set of normal and pathologic brain images by a team of neuroradiologists. The manual segmentation was performed using a human-computer interface specifically designed for pixel-by-pixel segmentation of brain images. The registration of corresponding images from different brains was accomplished using an elastic transformation. The presented segmentation method uses the multi-parameter model in adaptive segmentation of brain images on a pixel-by-pixel basis. The method was evaluated on a set of multi-parameter MR brain images of a twelve-year old patient 48h after suffering a stroke. The results of classification as compared to the manual segmentation of the same data show the efficacy and accuracy of the presented methods as well as its capability to create and learn new tissue classes.

Volumetric measurement of multifocal brain lesions. Implications for treatment trials of vascular dementia and multiple sclerosis.

This pilot study examined the reproducibility of serial magnetic resonance (MR) measurements of brain, ventricular, sulcal, and lesion volumes in patients with ischemic brain disease using an image analysis protocol designed at the University of Cincinnati. Five patients with a clinical history of brain ischemia had two separate MR brain imaging studies using the standard clinical MR imaging protocol at the University of Cincinnati Medical Center. The MR images on both film and tape were digitized and then analyzed according to the standardized image analysis protocol. Based on tape data, variability in volume measurements between the two MR studies, as measured by the coefficient of variation, ranged from 1% for intracranial volume to 8% for ventricular volume. Variability based on film data was slightly greater, ranging from 2% for intracranial volume to 12% for lesion volume. As part of a multicenter treatment trial of vascular dementia, this method was then used to analyze MR films in 13 patients with vascular dementia who all had an MR study at baseline and at 1 year. The mean annual change in lesion volume was 4 +/- 5 cm3 (a 24% increase from the baseline lesion volume); in ventricular volume, 7 +/- 8 cm3 (a 10% increase from baseline); and in sulcal volume, 13 +/- 25 cm3 (a 5% increase from baseline). This method of image analysis, using MR film or tape-generated data, can provide reproducible serial measurements of brain, ventricular, sulcal, and ischemic lesion volumes. This method, if applied in randomized treatment trials of vascular dementia or multiple sclerosis, can be used to monitor disease progression and to evaluate the effectiveness of a given therapy.

Analysis of mammographic microcalcifications using gray-level image structure features.

At present, mammography associated with clinical breast examination and breast self-examination is the only effective and viable method for mass breast screening. The presence of microcalcifications is one of the primary signs of breast cancer. It is, difficult however, to distinguish between benign and malignant microcalcifications associated with breast cancer. Here, the authors define a set of image structure features for classification of malignancy. Two categories of correlated gray-level image structure features are defined for classification of "difficult-to-diagnose" cases. The first category of features includes second-order histogram statistics-based features representing the global texture and the wavelet decomposition-based features representing the local texture of the microcalcification area of interest. The second category of features represents the first-order gray-level histogram-based statistics of the segmented microcalcification regions and the size, number, and distance features of the segmented microcalcification cluster. Various features in each category were correlated with the biopsy examination results of 191 "difficult-to-diagnose" cases for selection of the best set of features representing the complete gray-level image structure information. The selection of the best features was performed using the multivariate cluster analysis as well as a genetic algorithm (GA)-based search method. The selected features were used for classification using backpropagation neural network and parameteric statistical classifiers. Receiver operating characteristic (ROC) analysis was performed to compare the neural network-based classification with linear and k-nearest neighbor (KNN) classifiers. The neural network classifier yielded better results using the combined set of features selected through the GA-based search method for classification of "difficult-to-diagnose" microcalcifications.

Three-dimensional anatomical model-based segmentation of MR brain images through Principal Axes Registration.

Model-based segmentation and analysis of brain images depends on anatomical knowledge which may be derived from conventional atlases. Classical anatomical atlases are based on the rigid spatial distribution provided by a single cadaver. Their use to segment internal anatomical brain structures in a high-resolution MR brain image does not provide any knowledge about the subject variability, and therefore they are not very efficient in analysis. We present a method to develop three-dimensional computerized composite models of brain structures to build a computerized anatomical atlas. The composite models are developed using the real MR brain images of human subjects which are registered through the Principal Axes Transformation. The composite models provide probabilistic spatial distributions, which represent the variability of brain structures and can be easily updated for additional subjects. We demonstrate the use of such a composite model of ventricular structure to help segmentation of the ventricles and Cerebrospinal Fluid (CSF) of MR brain images. In this paper, a composite model of ventricles using a set of 22 human subjects is developed and used in a model-based segmentation of ventricles, sulci, and white matter lesions. To illustrate the clinical usefulness, automatic volumetric measurements on ventricular size and cortical atrophy for an additional eight alcoholics and 10 normal subjects were made. The volumetric quantitative results indicated regional brain atrophy in chronic alcoholics.

Iterative Principal Axes Registration method for analysis of MR-PET brain images.

Computerized automatic registration of MR-PET images of the brain is of significant interest for multimodality brain image analysis. In this paper, we discuss the Principal Axes Transformation for registration of three-dimensional MR and PET images. A new brain phantom designed to test MR-PET registration accuracy determines that the Principal Axes Registration method is accurate to within an average of 1.37 mm with a standard deviation of 0.78 mm. Often the PET scans are not complete in the sense that the PET volume does not match the respective MR volume. We have developed an Iterative Principal Axes Registration (IPAR) algorithm for such cases. Partial volumes of PET can be accurately registered to the complete MR volume using the new iterative algorithm. The quantitative and qualitative analyses of MR-PET image registration are presented and discussed. Results show that the new Principal Axes Registration algorithm is accurate and practical in MR-PET correlation studies.

3-D image analysis of intra-cerebral brain hemorrhage from digitized CT films.

A new 3-D technique for the segmentation and quantification of human spontaneous intra-cerebral brain hemorrhage (ICH) is presented in this paper. The algorithm for ICH primary region segmentation uses the spatially weighted K-means histogram-based clustering algorithm. The ICH edema region segmentation algorithm employs an iterative morphological processing of the ICH brain data. A volume rendering technique is used for the effective 3-D visualization of ICH segmented regions. A computer program is developed for use in the human spontaneous ICH study involving a large number of patients. Experimental measurements and visualization results are presented which were computed on real ICH patient brain data.

A gray-level thinning method for delineation and representation of arteries.

The quantification of three dimensional (3D) properties of coronary arteries is of significant importance. The performance of the 3D analysis is critically based on low-level representation of the arterial tree for different projections. A skeletal representation of arteries can provide appropriate data structure for registration of multiple angiographic projections and it can be further utilized for 3D reconstruction of the arterial tree. This paper presents an automated method for extracting the skeletal points of an arterial tree directly from the gray-level information without determining the edges a priori. It offers the advantage of improved reliability compared to methods based on detecting dual edges of the arteries. Novel application of filtering techniques provide accurate estimates of the statistics of the background. A recursive search scheme is used to aggregate the skeletal representation at multiple resolutions. Results on a set of Digitally Subtracted Angiograms (DSA) have been presented.

Noninvasive imaging and analysis of skin lesions for early detection of cutaneous malignant melanoma.

Knowledge of the depth of invasion and three-dimensional (3D) characteristics of skin lesions is crucial to the early diagnosis and prognosis of malignant melanoma. These parameters are currently available only after biopsy. A clinical prototype of an optical imaging instrument, called a nevoscope, has been developed to noninvasively image skin lesions. The nevoscope provides multiple views of a transilluminated skin lesion. These views provide a three-dimensional perception of the lesion. In addition, image-analysis techniques are applied to these images to quantify various diagnostic parameters, such as 3D geometric features and surface texture and pigmentation characteristics. Clinical trials are under way to evaluate the utility of the nevoscope as a diagnostic tool for early detection of melanoma. The construction of the clinical prototype and the integrated clinical imaging environment are described. The development of iterative reconstruction techniques suitable for nevoscopy data is also described. Experimental reconstructions of a physical tissue-like phantom and an in-vivo lesion are presented. The phantom is recovered effectively. Thus, nevoscopy is of potential diagnostic value.

Segmentation of medical images through competitive learning.

In image analysis applications, segmentation of gray-level images into meaningful regions is an important low-level processing step. Various approaches to segmentation investigated in the literature, in general, use either local information of gray-level values of pixels (region growing based methods, for example) or the global information (histogram thresholding based methods, for example). Application of these approaches for segmenting medical images often does not provide satisfactory results. Medical images are usually characterized by low local contrast and noisy or faded features causing unacceptable performance of local information based segmentation methods. In addition, because of a large amount of structural information found in medical images, global information based segmentation methods yield inadequate results in region extraction. We present a novel approach to image segmentation that combines local contrast as well as global gray-level distribution information. The presented method adaptively learns useful features and regions through the use of a normalized contrast function as a measure of local information and a competitive learning based method to update region segmentation incorporating global information about the gray-level distribution of the image. In this paper, we present the framework of such a self organizing feature map, and show the results on simulated as well as real medical images.

Functional importance of ventricular enlargement and cortical atrophy in healthy subjects and alcoholics as assessed with PET, MR imaging, and neuropsychologic testing.

The authors assessed the relationship between ventricular enlargement, cortical atrophy, regional brain glucose metabolism, and neuropsychologic performance in 10 alcoholics and 10 control subjects. Regional brain glucose metabolism was measured with fluorine-18 fluorodeoxyglucose (FDG) and positron emission tomography (PET). Cortical atrophy and ventricular size were evaluated quantitatively with magnetic resonance (MR) imaging. Alcoholics had decreased brain glucose metabolism and more cortical atrophy but did not have significantly greater ventricular size than did control subjects. The degree of ventricular enlargement and of cortical atrophy was associated with decreased metabolism predominantly in the frontal cortices and subcortical structures in both alcoholics and control subjects. There were no significant correlations between neuropsychologic performance and MR imaging structural changes, whereas various subtest scores were significantly correlated with frontal lobe metabolism. These data show that F-18 FDG PET is a sensitive technique for detecting early functional changes in the brain due to alcohol and/or aging before structural changes can be detected with MR imaging.

Knowledge-based multi-modality three-dimensional image analysis of the brain.

With the recent advances in medical imaging, three-dimensional anatomical and metabolic images of the brain are now available through MR/CT and PET/SPECT imaging modalities. Computerized multi-modality three-dimensional brain image registration and analysis can provide important correlated information for improving diagnosis and studying the pathology of disease. Such analysis may also provide help in planning brain surgery. Further, an anatomical model based quantification and analysis of internal structure can be used to develop a computerized anatomical atlas. Conventional anatomical atlases provide rigid spatial distribution of internal structures extracted from a single subject. The proposed computerized anatomical atlas provides probabilistic spatial distributions which can be easily updated to incorporate the variability of brain structures of subjects selected from pre-defined groups. This paper first presents a review of the current trends in knowledge-based segmentation, labeling, and analysis of MR brain images and then describes the Principal Axes Transformation based registration of three-dimensional MR brain images to develop composite models of selected internal brain structures. The composite models can be used as a computerized anatomical atlas in model-based segmentation and labeling of MR brain images. Three-dimensional labeled MR images of the brain can also be registered and correlated with PET images for analyzing the metabolic activity in the anatomically selected volume of interest. On the other hand, a volume of interest can be selected using the metabolic information and then analyzed for correlated anatomical information using the registered MR-PET images.

Three-dimensional imaging and reconstruction of skin lesions.

An optical instrument called a "nevoscope" is used to image skin lesions by transillumination with visible light. The lesion is transilluminated by a fiber-optic annular ring light source that directs light into the skin area surrounding the lesion, forming a virtual source just beneath the lesion. Mirrors uniformly spaced around the lesion and tilted at various angles provide orthographic projections of the skin lesion. Additional views are obtained by rotating the mirror assembly. These multiple views are used in a direct three-dimensional (3D) reconstruction of the lesion using a filtered backprojection method. In this paper, we discuss the methodology of direct 3D reconstruction from 2D views of a transilluminated skin lesion as obtained using the new prototype nevoscope. We present the results of direct 3D reconstruction of a simulated phantom and a test object imaged using the nevoscope. In addition, a skin lesion was scanned in situ using the new prototype nevoscope. Results of the reconstruction of this lesion are also presented.

Segmentation of images of skin lesions using color and texture information of surface pigmentation.

Image segmentation algorithms extract regions on the basis of similarity of a predefined image feature such as gray-level value. In many applications, images that exhibit a variety of structure or texture cannot be adequately segmented by gray-level values alone. Additional features related to the structure of the image are needed to segment such images. Images of skin lesions exhibit significant variations in color hues as well as geometrical appearance of local surface structure. For example, images of cutaneous malignant melanoma exhibit a rich combination of color and geometrical structure of pigmentation. In these images, the local repetition of the geometrical surface structure provides the basis for the appearance of a texture pattern in the neighborhood region. For obtaining meaningful segmentation of images of skin lesions, a multichannel segmentation algorithm is proposed in this paper which uses both gray-level intensity and texture-based features for region extraction. The intensity-based segmentation is obtained using the modified pyramid-based region extraction algorithm. The texture-based segmentation is obtained by a bilevel shifted-window processing algorithm that uses new generalized co-occurrence matrices. The results of individual segmentations obtained from different channels, representing the complete set of color and texture information, are analyzed using heuristic merging rules to obtain the final color- and texture-based segmentation. Simulated as well as real images of skin lesions, representing various color shades and textures, have been processed. We show that using contrast link information in the pyramid-based region extraction process, and using the absolute magnitude and directional information in the generalized co-occurrence matrices (GCM) method, significant improvement in image segmentation can be obtained. Further, by incorporating the merging rules better results are obtained than those obtained using the gray-level intensity feature alone.

Knowledge-based 3D analysis from 2D medical images.

An anatomical knowledge-based system for image analysis that interprets CT/MR (computed tomography/magnetic resonance) images of the human chest cavity is reported. The approach utilizes a low-level image analysis system with the ability to analyze the data in bottom-up (or data-driven) and top-down (or model-driven) modes to improve the high-level recognition process. Several image segmentation algorithms, including K-means clustering, pyramid-based region extraction, and rule-based merging, are used for obtaining the segmented regions. To obtain a reasonable number of well-segmented regions that have a good correlation with the anatomy, a priori knowledge in the form of masks is used to guide the segmentation process. Segmentation of the brain is also considered.

Knowledge-based analysis and understanding of medical images.

Knowledge-based image analysis and interpretation of radiological images is of significant interest for several reasons including a means to identify and label each part of the image for further automated diagnostic analysis. Also, there is a need to develop a knowledge-based biomedical image analysis system which can analyze and interpret the anatomical images (such as those obtained from X-ray computed tomography (CT) scanning) in order to help analysis of functional images (such as those obtained from positron emission tomography (PET) scanning) of the organ of the same patient. This paper deals with the design and implementation of a knowledge-based system to analyze and interpret CT anatomical images of the human chest. In the approach presented here, the emphasis has been on the development of a strong low-level analysis system with the capability of analyzing in both bottom-up and top-down modes; and on the use of hierarchical relational, spatial, and structural knowledge of human anatomy in the process of high-level analysis and recognition.

A review on biomedical image processing and future trends.

The last two decades have witnessed a revolutionary development in the field of biomedical and diagnostic imaging. Imaging procedures and modalities which were only in the experimental research phase in the early part of the last two decades, have now become universally accepted clinical procedures. They include computerized tomography (CT), magnetic resonance imaging, ultrasound imaging, nuclear medicine imaging, computerized hematological cell analysis, etc. In the past, the conventional and relatively simple image processing techniques such as image enhancement, gray-level mapping, spectral analysis, region extraction, etc. have been modified for biomedical images and successfully applied for processing and analysis. The role of image enhancement, gray-level mapping, and image reconstruction from projections algorithms in CT and other radiological imaging modalities is well evident. Recently, many advances in biomedical image processing, analysis, and understanding algorithms have shown a great potential for enhancing and interpreting useful diagnostic information from these images more accurately. This paper presents a review on the current state-of-the-art techniques in biomedical image processing and comments on future trends.