Set Of Standard Images Research Articles

Image reconstruction and the effect on dose calculation for hip prostheses

High atomic number inserts, such as hip prostheses and dental fillings, cause streak artifacts on computed tomography (CT) images when filtered back-projection (FBP) methods are used. These streak artifacts severely degrade our ability to differentiate the tumor volume. Also, incorrect Hounsfield numbers yield incorrect electron density information that may lead to erroneous dose calculations, and, as a result, compromise clinical outcomes. The aim of this research was to evaluate the dosimetric consequences of artifacts during radiotherapy planning of a prostate patient containing a hip prosthesis. The CT numbers corresponding to an iron prosthesis were inserted into the right femoral head of an existing CT image set. This artifact-free image was used as the standard image set. CT projections through the image set formed the sinogram, from which filtered back projection and iterative deblurring methods were used to create reconstructed image sets. These reconstructed image sets contained artifacts. Prostate treatment plans were then calculated using a Monte Carlo system for the standard and reconstructed CT image sets. Close to the prosthesis, the CT numbers between the reconstructed and standard image sets differed substantially. However, because the CT number differences covered only a small area, the dose distributions on the reconstructed and standard image sets were not significantly different. The dose-volume histograms for the prostate, rectum, and bladder were virtually identical. Our results indicate that even though CT image artifacts restrict our ability to differentiate tumors and critical structures, the dose distributions for a prostate plan containing a hip prosthesis, calculated on both artifact-free image sets and image sets containing artifacts, are not significantly different.

A multichannel watershed-based algorithm for supervised texture segmentation

Segmentation of image regions based on their texture is a standard problem in image analysis. Once a set of texture features is selected, several algorithms can be applied to segment the image into regions. This paper presents an extension of the watershed algorithm using a vector gradient and multivariate region merging methods. The algorithm uses a set of texture images, and it only depends on an adjustable parameter. Results are presented on a standard set of synthetic images and on textured medical ones, using different texture parameters and merging criteria.

Suboptimal minimum cluster volume cover-based method for measuring fractal dimension

A new method for calculating fractal dimension is developed in this paper. The method is based on the box dimension concept; however, it involves direct estimation of a suboptimal covering of the data set of interest. By finding a suboptimal cover, this method is better able to estimate the required number of covering elements for a given cover size than is the standard box counting algorithm. Moreover, any decrease in the error of the covering element count directly increases the accuracy of the fractal dimension estimation. In general, our method represents a mathematical dual to the standard box counting algorithm by not solving for the number of boxes used to cover a data set given the size of the box. Instead, the method chooses the number of covering elements and then proceeds to find the placement of smallest hyperellipsoids that fully covers the data set. This method involves a variant of the Fuzzy-C Means clustering algorithm, as well as the use of the Minimum Cluster Volume clustering algorithm. A variety of fractal dimension estimators using this suboptimal covering method are discussed. Finally, these methods are compared to the standard box counting algorithm and wavelet-decomposition methods for calculating fractal dimension by using one-dimensional cantor dust sets and a set of standard Brownian random fractal images.

Comparative Study of Image Compression Techniques for Digital Particle Image Velocimetry

A comparative, quantitative study of image compression techniques for use with digital particle image velocimetry has been performed. Several candidate compression algorithms were selected for the study including a lossless technique and a series of mathematical transform-based methods. Each of the compression algorithms was implemented using commercial off-the-shelf software packages. Three image sequences were selected to exercise the various compression methods. These sequences included a set of industry standard images and two sets of images obtained from experimental work conducted at NASA Langley. Evaluation of the various methods was accomplished using quantitative perceptual and metrological performance measures. The results of the study indicate that several methods of compression are applicable to digital particle image velocimetry images. For selected applications where thresholding of the image gray levels can be performed, lossless algorithms yield excellent performance in terms of compression rates and introduction of negligible errors to the images. For applications where thresholding is not appropriate, JPEG and wavelet-based algorithms yield high compression rates with minor spatial errors introduced to the images. Fractal-based algorithms were shown to be capable of providing high compression rates, but the technique must be guided via the use of a suitably selected template image if spatial errors are to be minimized in the compressed images.

The Diagnosis of Small Solitary Pulmonary Nodule: Comparison of Standard and Inverse Digital Images on a High-Resolution Monitor using ROC Analysis

Purpose: To study the impact of inversion of soft-copy chest radiographs on the detection of small solitary pulmonary nodules using a high-resolution monitor. Materials and Methods: The study group consisted of 80 patients who had undergone posterior chest radiography; 40 had a solitary noncalcified pulmonary nodule approximately 1 cm in diameter, and 40 were control subjects. Standard and inverse digital images using the inversion tool on a PACS system were displayed on high-resolution monitors (2048x2560x8 bit). Ten radiologists were requested to rank each image using a five-point scale (1=definitely negative, 3=equivocal or indeterminate, 5=definite nodule), and the data were interpreted using receiver operating characteristic (ROC) analysis. Results: The area under the ROC curve for pooled data of standard image sets was significantly larger than that of inverse image sets (0.8893 and 0.8095, respectively; p 0.05). Conclusion: For detecting small solitary pulmonary nodules, inverse digital images were significantly inferior to standard digital images.

Multiple eigenspaces

In this paper, we propose a novel self-organizing framework to construct multiple, low-dimensional eigenspaces from a set of training images. Grouping of images is systematically and robustly performed via eigenspace-growing in terms of low-dimensional eigenspaces. To further increase the robustness, the eigenspace-growing is initiated independently with many small groups of images—seeds. All these grown eigenspaces are treated as hypotheses that are subject to a selection procedure eigenspace-selection, based on the MDL principle, which selects the final resulting set of eigenspaces as an efficient representation of the training set, taking into account the number of images encompassed by the eigenspaces, the dimensions of the eigenspaces, and their corresponding residual errors. We have tested the proposed method on a number of standard image sets, and the significance of the approach with respect to the recognition rate has been demonstrated.

Comparison of the multilayer perceptron with neuro-fuzzy techniques in the estimation of cover class mixture in remotely sensed data

Mixed pixels are a major source of inconvenience in the classification of remotely sensed data. This paper compares MLP with so-called neuro-fuzzy algorithms in the estimation of pixel component cover classes. Two neuro-fuzzy networks are selected from the literature as representatives of soft classifiers featuring different combinations of fuzzy set-theoretic principles with neural network learning mechanisms. These networks are: 1) the fuzzy multilayer perceptron (FMLP) and 2) a two-stage hybrid (TSH) learning neural network whose unsupervised first stage consists of the fully self-organizing simplified adaptive resonance theory (FOSART) clustering model, FMLP, TSH, and MLP are compared on CLASSITEST, a standard set of synthetic images where per-pixel proportions of cover class mixtures are known a priori. Results are assessed by means of evaluation tools specifically developed for the comparison of soft classifiers. Experimental results show that classification accuracies of FMLP and TSH are comparable, whereas TSH is faster to train than FMLP. On the other hand, FMLP and TSW outperform MLP when little prior knowledge is available for training the network, i.e., when no fuzzy training sites, describing intermediate label assignments, are available.

Using Renyi's Information and Wavelets for Target Detection: An Application to Mammograms

In this paper we present a multi-scale method for the detection of small targets embedded in noisy background. The multi-scale representation is built using a weighted undecimated discrete wavelet transform. The method, in essence, is based on the maximisation of information available at each resolution level of the representation. We show that such objective can be achieved by maximising Renyi’s information. This approach allows us to determine an adaptive threshold useful for discriminating, at each scale, between wavelet coefficients representing targets and those representing background noise. Eventually, avoiding inverse transformation, scale-dependent estimates are combined according to a majority vote strategy. The proposed technique is experimented on a standard data set of mammographic images.

Lossless image coding via adaptive linear prediction and classification

In past years, there have been several improvements in lossless image compression. All the recently proposed state-of-the-art lossless image compressors can be roughly divided into two categories: single and double-pass compressors. Linear prediction is rarely used in the first category, while TMW, a state-of-the-art double-pass image compressor, relies on linear prediction for its performance. We propose a single-pass adaptive algorithm that uses context classification and multiple linear predictors, locally optimized on a pixel-by-pixel basis. Locality is also exploited in the entropy coding of the prediction error. The results we obtained on a test set of several standard images are encouraging. On the average, our ALPC obtains a compression ratio comparable to CALIC while improving on some images.

Generalized hidden Markov models. II. Application to handwritten word recognition

For part I see ibid. vol.8, no. 1 (2000). This paper presents an application of the generalized hidden Markov models to handwritten word recognition. The system represents a word image as an ordered list of observation vectors by encoding features computed from each column in the given word image. Word models are formed by concatenating the state chains of the constituent character hidden Markov models. The novel work presented includes the preprocessing, feature extraction, and the application of the generalized hidden Markov models to handwritten word recognition. Methods for training the classical and generalized (fuzzy) models are described. Experiments were performed on a standard data set of handwritten word images obtained from the US Post Office mail stream, which contains real-word samples of different styles and qualities.

A two-layered wavelet-based algorithm for efficient lossless and lossy image compression

In this paper, we propose a wavelet-based image-coding scheme allowing lossless and lossy compression, simultaneously. Our two-layered approach utilizes the best of two worlds: it uses a highly performing wavelet-based or wavelet packet-based coding technique for lossy compression in the low bit range as a first stage. For the second (optional) stage, we extend the concept of reversible integer wavelet transforms to the more flexible class of adaptive reversible integer wavelet packet transforms which are based on the generation of a whole library of bases, from which the best representation for a given residue between the reconstructed lossy compressed image and the original image is chosen using a fast-search algorithm. We present experimental results demonstrating that our compression algorithm yields a rate-distortion performance similar or superior to the best currently published pure lossy still image-coding methods. At the same time, the lossless compression performance of our two-layered scheme is comparable to that of state-of-the-art pure lossless image-coding schemes. Compared to other combined lossy/lossless coding schemes such as the emerging JPEG-2000 still image-coding standard PSNR improvements up to 3 dB are achieved for a set of standard test images.

Capacity of full frame DCT image watermarks

The evaluation of the number of bits that can be hidden within an image through digital watermarking is a crucial topic, which has been addressed only for additive watermarks. The evaluation of watermark capacity is very important because it allows to put a theoretical upper bound on the amount of information that can be hidden into an image by a given watermarking procedure, regardless of the watermark extraction technique. It is the purpose of this work to suggest a methodology for the evaluation of the watermark capacity in a nonadditive, non-Gaussian framework, and to discuss the results we obtained by applying it to a set of standard images.

Post-Processing Algorithms of 3D MR Angiography in the Evaluation of Patients with Intracranial Aneurysms

Our purpose was to evaluate whether useful information was gained with 3D surface rendering and endoscopic visualisation post-processing techniques of MR angiographic data when compared to standard rotated MIP image sets in patients with intracranial aneurysms. We prospectively studied 49 patients with 55 intracranial aneurysms confirmed by digital subtraction angiography (DSA) using MR angiography performed on a 1.5-T system. Our protocol included 3D time-of-flight (TOF) with magnetisation transfer and 3D phase-contrast (PC) following contrast administration. Each image set consisted of either 28 or 60 slices. Post-processing on a SUN work-station with GE software included: MIP, 3D surface rendering, and “advantage navigator” software for endoluminal imaging. The sac, neck and parent arteries of all aneurysms were depicted on both MIP projection and 3D surface rendering images. Objective evaluation demonstrated that surface rendering delineated aneurysm direction, shape, neck size, and spatial relationship to parent vessels; the MIP method was unable to depict this information. 3D TOF acquisition was superior to 3D PC in small vessel visualisation due to implementation of higher spatial resolution resulting in isometric submillimeter voxels. Magnetisation transfer, selective reconstruction, real-time rotation, and subtraction of data further optimised imaging. Endoluminal reconstructions (possible in 52/55 aneurysms) were particularly useful in identifying parent artery incorporation by the aneurysm. Surface rendering and endoluminal reconstruction techniques provide useful information for therapy planning of intracranial aneurysms not provided by MIP projections. Objective analysis of the value of endoluminal 3D imaging regarding aneurysm and parent artery wall morphology awaits further optimisation of the technique.

Differential block coding of bilevel images

In this correspondence, a simple one-dimensional (1-D) differencing operation is applied to bilevel images prior to block coding to produce a sparse binary image that can be encoded efficiently using any of a number of well-known techniques. The difference image can be encoded more efficiently than the original bilevel image whenever the average run length of black pixels in the original image is greater than two. Compression is achieved because the correlation between adjacent pixels is reduced compared with the original image. The encoding/decoding operations are described and compression performance is presented for a set of standard bilevel images.

Positron emission tomography study of voluntary saccadic eye movements and spatial working memory.

1. The purpose of this study is to define the cortical regions that subserve voluntary saccadic eye movements and spatial working memory in humans. 2. Regional cerebral blood flow (rCBF) during performance of oculomotor tasks was measured with [15O]-H2O positron emission tomography (PET). Eleven well-trained, healthy young adults performed the following tasks: visual fixation, visually guided saccades, antisaccades (a task in which subjects made saccades away from rather than toward peripheral targets), and either an oculomotor delayed response (ODR, a task requiring memory-guided saccades after a delay period) or a conditional antisaccade task (a task in which the color of the peripheral target determined whether a saccade toward or away from the target was required). An additional six subjects performed a sequential hand movement task to compare localization of hand-related motor cortex and the frontal eye fields (FEFs) and of the hand- and eye-movement-related regions of the supplementary motor area (SMA). 3. Friston's statistical parametric mapping (SPM) method was used to identify significant changes in rCBF associated with task performance. Because SPM does not take advantage of the anatomic information available in magnetic resonance (MR) scans, each subject's PET scan was registered to that individual's MR scan, after which all PET and MR studies were transformed to conform to a standard reference MR image set. Subtraction images were visually inspected while overlayed on the reference MR scan to which PET images had been aligned, in order to confirm anatomic localization of significant rCBF changes. 4. Compared with visual fixation, performing visually guided saccades led to a significant bilateral activation in FEF, cerebellum, striate cortex, and posterior temporal cortex. Right posterior thalamus activation was also observed. 5. The visually guided saccade task served as the comparison task for the ODR, antisaccade, and conditional antisaccade tasks for identification of task-related changes in rCBF beyond those associated with saccade execution. Performance on the ODR task was associated with a bilateral increase of rCBF in FEFs, SMA, dorsolateral prefrontal cortex (DLPFC), and posterior parietal cortex. The cortical regions of increased regional blood flow during the ODR task also showed increased rCBF during the antisaccade task; however, FEF and SMA activations were significant only in the right hemisphere. These findings closely parallel those of single-cell recording studies with behaving monkeys in indicating that FEF, DLPFC, SMA, and posterior parietal cortex perform computational activity for voluntary purposive saccades. 6. Comparison of PET scans obtained during performance of eye movement and hand movement tasks indicated that peak activations in FEF were located approximately 2 cm lateral and 1 cm anterior to those of hand-related motor cortex. The oculomotor area of SMA, the supplementary eye field (SEF), was located approximately 7-8 mm anterior and superior to the hand-related area of SMA. 7. During performance of antisaccade and ODR tasks, rCBF was significantly lower in ventromedial prefrontal cortex (PFC), along the rectus gyrus, and in ventral anterior cingulate cortex than during the visually guided saccade and fixation tasks. During the antisaccade task, the ventral region of lower rCBF involved medial structures including left ventral striatum and bilateral medial temporal-limbic cortex. During the ODR task, the ventral aspect of the region of lower rCBF extended laterally, rather than medially, to include the temporal poles. The lower blood flow observed in ventromedial PFC during both the antisaccade and ODR tasks, relative to the visually guided saccade and fixation tasks, suggests that modulation of output from ventromedial PFC to limbic cortex and the striatum may play a role in the voluntary control of saccadic eye movements, possibly in the suppression of responses that would interrupt

Uses and limitations of spoiled gradient-refocused imaging in the evaluation of suspected intracranial tumors

This article describes the use of a radiofrequency-spoiled gradient-recalled (SPGR) imaging pulse sequence in the evaluation of intracranial masses. This pulse sequence provides excellent anatomic detail with T1-weighted image contrast. Rapid, sequential, single-slice (two-dimensional) images of the brain can be obtained in patients who are unable to hold still for long periods of time. In addition, volumetric (three-dimensional) image data sets can be obtained that provide extremely thin (1- to 2-mm) sections of high detail and good signal-to-noise ratio for selected critical structures within the brain. Finally, because SPGR is also utilized for time-of-flight angiography, parenchymal information can be obtained simultaneously with cerebral blood vessel definition. One potential pitfall that the magnetic resonance radiologist must be aware of is the fact that, in many patients, the degree of contrast enhancement is greatly diminished on postgadolinium SPGR images compared with conventional spin-echo T1-weighted images. Comparison with a set of standard spin-echo postcontrast images or, potentially, the use of higher doses of gadolinium may solve this problem. In spite of this limitation, the selective utilization of SPGR imaging can yield additional useful information for evaluation and preoperative planning in patients with intracranial masses.

Technique for the use of standard outlines for attenuation correction and quantification in SPECT

A technique is described for using a set of standard outline images for attenuation correction and quantification in SPECT. The effectiveness of the method was investigated using an anatomical phantom and in a series of 123I-mIBG patient imaging studies. In the latter measurements the ability of the technique to quantify the activity of various features in the images was compared to the use of CT images for attenuation correction. The mean difference between the two measurements was 4.4%.