Evaluation of image reconstruction methods for (123)I-MIBG-SPECT: a rank-order study.

BACKGROUND: There is an opportunity to improve the image quality and lesion detectability in single photon emission computed tomography (SPECT) by choosing an appropriate reconstruction method and optimal parameters for the reconstruction. PURPOSE: To optimize the use of the Flash 3D reconstruction algorithm in terms of equivalent iteration (EI) number (number of subsets times the number of iterations) and to compare with two recently developed reconstruction algorithms ReSPECT and orthogonal polynomial expansion on disc (OPED) for application on (123)I-metaiodobenzylguanidine (MIBG)-SPECT. MATERIAL AND METHODS: Eleven adult patients underwent SPECT 4 h and 14 patients 24 h after injection of approximately 200 MBq (123)I-MIBG using a Siemens Symbia T6 SPECT/CT. Images were reconstructed from raw data using the Flash 3D algorithm at eight different EI numbers. The images were ranked by three experienced nuclear medicine physicians according to their overall impression of the image quality. The obtained optimal images were then compared in one further visual comparison with images reconstructed using the ReSPECT and OPED algorithms. RESULTS: The optimal EI number for Flash 3D was determined to be 32 for acquisition 4 h and 24 h after injection. The average rank order (best first) for the different reconstructions for acquisition after 4 h was: Flash 3D(32) > ReSPECT > Flash 3D(64) > OPED, and after 24 h: Flash 3D(16) > ReSPECT > Flash 3D(32) > OPED. A fair level of inter-observer agreement concerning optimal EI number and reconstruction algorithm was obtained, which may be explained by the different individual preferences of what is appropriate image quality. CONCLUSION: Using Siemens Symbia T6 SPECT/CT and specified acquisition parameters, Flash 3D(32) (4 h) and Flash 3D(16) (24 h), followed by ReSPECT, were assessed to be the preferable reconstruction algorithms in visual assessment of (123)I-MIBG images.

Allelic loss of chromosomes 8 and 19 in MENX-associated rat pheochromocytoma.

Pheochromocytomas are neoplasias of neural crest origin that arise from the chromaffin cells of the adrenal medulla. Pheochromocytomas arise with complete penetrance in rats homozygous for a germ-line frameshift mutation of Cdkn1b, encoding the cell cycle inhibitor p27KIP1 (MENX syndrome). We performed a genome-wide scan for allelic imbalance comparing 20 rat pheochromocytoma DNAs with normal rat DNA to better understand the pathobiology of the tumors and to correlate the findings with human pheochromocytoma. We identified allelic imbalance (AI) at candidate regions on rat chromosomes 8 and 19. Interestingly, the regions often lost in rat tumors are syntenic to regions involved in human pheochromocytomas. Fluorescence in situ hybridization analysis further validated the AI data. Sdhd and Rassf1a were analyzed in detail as they map to regions of AI on chromosome 8 and their homologues are implicated in human pheochromocytoma: we found no genetic mutations nor decreased expression. We also analyzed additional candidate genes, that is, rat homologues of genes predisposing to human pheochromocytoma and known tumor-suppressor genes, but we found no AI. In contrast, we observed frequent overexpression of Cdkn2a and Cdkn2c, encoding the cell cycle inhibitors p16INK4a and p18INK4c, respectively. The relative small number of allelic changes we found in rat pheochromocytoma might be related to their nonmalignant status and losses at chromosomes 8 and 19 are events that precede malignancy. Because of the high concordance of affected loci between rat and human tumors, studies of the MENX-associated pheochromocytomas should facilitate the identification of novel candidate genes implicated in their human counterpart.

Comparison of interpolation functions to improve a rebinning-free CT-reconstruction algorithm.

The robust algorithm OPED for the reconstruction of images from Radon data has been recently developed. This reconstructs an image from parallel data within a special scanning geometry that does not need rebinning but only a simple re-ordering, so that the acquired fan data can be used directly for the reconstruction. However, if the number of rays per fan view is increased, there appear empty cells in the sinogram. These cells need to be filled by interpolation before the reconstruction can be carried out. The present paper analyzes linear interpolation, cubic splines and parametric (or "damped") splines for the interpolation task. The reconstruction accuracy in the resulting images was measured by the Normalized Mean Square Error (NMSE), the Hilbert Angle, and the Mean Relative Error. The spatial resolution was measured by the Modulation Transfer Function (MTF). Cubic splines were confirmed to be the most recommendable method. The reconstructed images resulting from cubic spline interpolation show a significantly lower NMSE than the ones from linear interpolation and have the largest MTF for all frequencies. Parametric splines proved to be advantageous only for small sinograms (below 50 fan views).

Reduction of anatomical noise in medical X-ray images.

The X-ray pattern of a mass of very fine non-distinguishable anatomical structures alters completely from radiograph to radiograph due to the unavoidable movements of the patient during the exposure. The corresponding image component shows noise-like behaviour and is therefore referred to as the anatomical noise. Reducing this component would enhance the quality of the clinical X-ray image and increase the detectability of radiological signal. We have found that by comparing two X-ray images of the same anatomy acquired under slightly different imaging geometry, it is possible to reduce the anatomical noise in one of the image pair. The proposed method, which allows this, is based on the appropriate attenuation in the wavelet domain. The values of attenuating factors for the wavelet coefficients are proportional to the correlation between the corresponding features of both images. This method was tested for different changes in the imaging geometry. In the case of no geometrical changes, only the quantum and the electronic noise are reduced. An effect of de-noising for the investigated images is obvious.

Comparison of technical and anatomical noise in digital thorax X-ray images.

Former studies by Hoeschen and Buhr indicated a higher total noise in a thorax image than expected from technical noise, i.e. quantum and detector noise. This difference results from the overlay of many small anatomical structures along the X-ray beam, which leads to a noise-like appearance without distinguishable structures in the projected image. A method is proposed to quantitatively determine this 'anatomical noise' component, which is not to be confused with the anatomical background (e.g. ribs). This specific anatomical noise pattern in a radiograph changes completely when the imaging geometry changes because different small anatomical structures contribute to the projected image. Therefore, two images are taken using slightly different exposure geometry, and a correlation analysis based on wavelet transforms allows to determining the uncorrelated noise components. Since the technical noise also differs from image to image, which makes it difficult to separate the anatomical noise, images of a lung phantom were produced on a low-sensitive industrial X-ray film using high-exposure levels. From these results, the anatomical noise level in real clinical thorax radiographs using realistic exposure levels is predicted using the general dose dependence described in the paper text and compared with the quantum and detector noise level of an indirect flat-panel detector system. For consistency testing, the same lung phantom was imaged with the same digital flat-panel detector and the total image noise including anatomical noise is determined. The results show that the relative portion of anatomical noise may exceed the technical noise level. Anatomical noise is an important contributor to the total image noise and, therefore, impedes the recognition of anatomical structures.

Evaluation of a novel method of noise reduction using computer-simulated mammograms.

A novel method of noise reduction has been tested for mammography using computer-simulated images for which the truth is known exactly. This method is based on comparing two images. The images are compared at different scales, using a cross-correlation function as a measure of similarity to define the image modifications in the wavelet domain. The computer-simulated images were calculated for noise-free primary radiation using a quasi-realistic voxel phantom. Two images corresponding to slightly different geometry were produced. Gaussian noise was added with certain properties to simulate quantum noise. The added noise could be reduced by >70% using the proposed method without any noticeable corruption of the structures. It is possible to save 50% dose in mammography by producing two images (each 25% of the dose for a standard mammogram). Additionally, a reduction of the anatomical noise and, therefore, better detection rates of breast cancer in mammography are possible.

Nodule detection in digital chest radiography: effect of anatomical noise.

The image background resulting from imaged anatomy can be divided into those components that are meaningful to the observers, in the sense that they are recognised as separate structures, and those that are not. These latter components (reffered to as anatomical noise) can be removed using a method developed within the RADIUS group. The aim of the present study was to investigate whether the removal of the anatomical noise results in images where lung nodules with lower contrast can be detected. A receiver operating characteristic (ROC) study was therefore conducted using two types of images: clinical chest images and chest images in which the anatomical noise had been removed. Simulated designer nodules with a full-width-at-fifth-maximum of 10 mm but with varying contrast were added to the images. The contrast needed to obtain an area under the ROC curve of 0.80, C0.8, was used as a measure of detectability (a low value of C0.8 represents a high detectability). Five regions of the chest X ray were investigated and it was found that in all regions the removal of anatomical noise led to images with lower C0.8 than the original images. On average, C0.8 was 20% higher in the original images, ranging from 7% (the lateral pulmonary regions) to 41% (the upper mediastinal regions).