Andrew Todd-Pokropek

Atlas-Based Segmentation of Degenerated Lumbar Intervertebral Discs From MR Images of the Spine

Intervertebral disc degeneration is an age-associated condition related to chronic back pain, while its consequences are responsible for over 90% of spine surgical procedures. In clinical practice, MRI is the modality of reference for diagnosing disc degeneration. In this study, we worked toward 2-D semiautomatic segmentation of both normal and degenerated lumbar intervertebral discs from T2-weighted midsagittal MR images of the spine. This task is challenged by partial volume effects and overlapping gray-level values between neighboring tissue classes. To overcome these problems three variations of atlas-based segmentation using a probabilistic atlas of the intervertebral disc were developed and their accuracies were quantitatively evaluated against manually segmented data. The best overall performance, when considering the tradeoff between segmentation accuracy and time efficiency, was accomplished by the atlas-robust-fuzzy c-means approach, which combines prior anatomical knowledge by means of a rigidly registered probabilistic disc atlas with fuzzy clustering techniques incorporating smoothness constraints. The dice similarity indexes of this method were 91.6% for normal and 87.2% for degenerated discs. Research in progress utilizes the proposed approach as part of a computer-aided diagnosis system for quantification and characterization of disc degeneration severity. Moreover, this approach could be exploited in computer-assisted spine surgery.

Scatter correction in scintigraphy : the state of the art

In scintigraphy, the detection of scattered photons degrades both visual image analysis and quantitative accuracy. Many methods have been proposed and are still under investigation to cope with scattered photons. The main features of the problem of scattering in radionuclide imaging are presented first, to provide a sound foundation for a critical review of the existing scatter correction techniques. These are described using a classification relating to their aims and principles. Their theoretical potentials are analysed, as well as the difficulties of their practical implementation. Finally, the problems of their evaluation and comparison are discussed.

7-Ketocholesterol favors lipid accumulation and colocalizes with Nile Red positive cytoplasmic structures formed during 7-ketocholesterol–induced apoptosis: Analysis by flow cytometry, FRET biphoton spectral imaging microscopy, and subcellular fractionation

Oxidized low‐density lipoproteins play key roles in atherosclerosis. Their toxicity is at least in part due to 7‐ketocholesterol (7KC), which is a potent inducer of apoptosis. In this study on human promonocytic U937 cells, we determined the effects and the interactions of 7KC with cellular lipids during 7KC‐induced apoptosis.

Scatter and cross-talk corrections in simultaneous Tc-99m/I-123 brain SPECT using constrained factor analysis and artificial neural networks

Simultaneous imaging of Tc-99m and I-123 would have a high clinical potential in the assessment of brain perfusion (Tc-99m) and neurotransmission (I-123) but is hindered by cross-talk between the two radionuclides. Monte Carlo simulations of 15 different dual-isotope studies were performed using a digital brain phantom. Several physiologic Tc-99m and I-123 uptake patterns were modeled in the brain structures. Two methods were considered to correct for cross-talk from both scattered and unscattered photons: constrained spectral factor analysis (SFA) and artificial neural networks (ANN). The accuracy and precision of reconstructed pixel values within several brain structures were compared to those obtained with an energy windowing method (WSA). In I-123 images, mean bias was close to 10% in all structures for SFA and ANN and between 14% (in the caudate nucleus) and 25% (in the cerebellum) for WSA. Tc-99m activity was overestimated by 35% in the cortex and 53% in the caudate nucleus with WSA, but by less than 9% in all structures with SFA and ANN. SFA and ANN performed well even in the presence of high-energy I-123 photons. The accuracy was greatly improved by incorporating the contamination into the SFA model or in the learning phase for ANN. SFA and ANN are promising approaches to correct for cross-talk in simultaneous Tc-99m/I-123 SPECT.

Clinical diagnosis from digital displays: Preliminary findings of the St Mary's evaluation project

Image quality is a fundamental issue in the introduction of picture archiving and communications systems (PACS), and one that has hitherto been eclipsed by other aspects of the considerable technological challenge facing scientists and manufacturers involved in its development. We conducted a formal evaluation of clinical radiological diagnosis from a commercially available PACS viewing station, using subperiosteal resorbtion in renal osteodystrophy as the test pathological diagnosis, with receiver operating characteristic (ROC) analysis of the results. We conclude that the displayed, digitised images were inferior to film using the apparatus tested and believe that careful, objective clinical evaluation of such systems is of paramount important.

Using an adaptive semiautomated self-evaluated registration technique to analyze MRI data for myocardial perfusion assessment.

To validate the adaptive semiautomated self‐evaluated registration technique (ASSERT) followed by factor analysis of medical image sequence (FAMIS) for analyzing myocardial perfusion using magnetic resonance imaging (MRI) images.

Experimental comparison of data transformation procedures for analysis of principal components

Results of principal component analysis depend on data scaling. Recently, based on theoretical considerations, several data transformation procedures have been suggested in order to improve the performance of principal component analysis of image data with respect to the optimum separation of signal and noise. The aim of this study was to test some of those suggestions, and to compare several procedures for data transformation in analysis of principal components experimentally. The experiment was performed with simulated data and the performance of individual procedures was compared using the non-parametric Friedmans test. The optimum scaling found was that which unifies the variance of noise in the observed images. In data with a Poisson distribution, the optimum scaling was the norm used in correspondence analysis. Scaling mainly affected the definition of the signal space. Once the dimension of the signal space was known, the differences in error of data and signal reproduction were small. The choice of data transformation depends on the amount of available prior knowledge (level of noise in individual images, number of components, etc), on the type of noise distribution (Gaussian, uniform, Poisson, other), and on the purpose of analysis (data compression, filtration, feature extraction).

Adaptive and self-evaluating registration method for myocardial perfusion assessment

With the advent of ultra-fast MR I, it is now possible to assess non-invasively regional myocardial perfusion with multislice coverage and sub-second temporal resolution. First-pass contrast enhanced studies are acquired with ECG-triggering and breath holding. Nevertheless, some respiratory induced movements still remain. Myocardial perfusion can be assessed locally byparametric imaging methods such as Factor Analysis of Medical Image Sequences (FAMIS), provided that residual motion can be corrected. An a posteriori registration method implemented in the image domain is proposed. It is based on an adaptive registration model of the heart combining three elementary shapes (left ventricle, right ventricle and pericardium). The registration procedure is performed on a potential map derived from the distance map. To evaluate the quality of the registration procedure a superimposition score between the registration model and the contour automatically extracted in the sequence is proposed. Rigid transformation hypotheses and registration analysis provide an efficient and automatic method which allows the rejection of outlier images, such as; outof synchronisation images, out of plane acquisitions. When compared to a manual registration method, this approach reduces processing time and requires a minimal intervention from the operator. The proposed method performs registration with a subpixel accuracy. It has been successfully applied to simulated images and clinical data. It should facilitate the use of MR first-pass perfusion studies in clinical practice.

Improved estimation of velocity and flow rate using regularized three‐point phase‐contrast velocimetry

We improved the three‐point phase‐contrast method by regularization of MR velocity data after acquisition of a low velocity‐to‐noise ratio (VNR) velocity image and a high VNR aliased velocity image. The phase unwrapping algorithm is based on the assumed correlation of the velocity of adjacent flow voxels on the low VNR and the unaliased high VNR images. We used Fourier encoding with eight velocity‐encoding gradient steps to obtain reference velocity images of the aorta from five subjects (274 images) and compared them with the phase‐contrast and three‐point phase‐contrast images with and without regularization. The VNR of the regularized velocity image was improved by 9.1 dB and the VNR of the three‐point phase‐contrast velocity image was improved by 0.7 dB with respect to the low first moment velocity image. Corresponding improvements of 9 dB and 3.7 dB were obtained for the estimations of instantaneous flow rate. Magn Reson Med 44:122–128, 2000.

Spatial regularization applied to factor analysis of medical image sequences (FAMIS)

Dynamic image sequences allow physiological mechanisms to be monitored after the injection of a tracer. Factor analysis of medical image sequences (FAMIS) hence creates a synthesis of the information in one image sequence. It estimates a limited number of structures (factor images) assuming that the tracer kinetics (factors) are similar at each point inside the structure. A spatial regularization method for computing factor images (REG-FAMIS) is proposed to remove irregularities due to noise in the original data while preserving discontinuities between structures. REG-FAMIS has been applied to two sets of simulations: (a) dynamic data with Gaussian noise and (b) dynamic studies in emission tomography (PET or SPECT), which respect real tomographic acquisition parameters and noise characteristics. Optimal regularization parameters are estimated in order to minimize the distance between reference images and regularized factor images. Compared with conventional factor images, the root mean square error between regularized images and reference factor images is improved by 3 for the first set of simulations, and by about 1.5 for the second set of simulations. In all cases, regularized factor images are qualitatively and quantitatively improved.