Diagnostic effects of edge sharpening filtration and magnification on digitally subtracted renal images

The improved appearance of digital radiographs filtered to improve local contrast and sharpen edges has not increased acceptance of these images by radiologists. Furthermore, many radiologists assert that correct diagnosis is not improved with these filtered images. This study was designed to test this assertion for digital subtraction angiograms (DSA) of renal images. Four experiments are described. First, phantom studies identified filters and their parameters thought likely to be acceptable and useful in diagnosing renal images formed by DSA. Second, these filters and parameters were then tested on medical images to assess their acceptance by radiologists. Third, display modes of windowing, positive/negative presentation, and magnification were varied for filtered and unfiltered images to assess preferences of radiologists. Fourth, filtered and unfiltered magnified images were used to test improved diagnosis. In the final experiment, 148 images from 33 renal studies (15 normal, 18 abnormal) were magnified, gray level windowed, and filtered. Diagnosis was not improved by the two edge sharpening filters tested.     

A method of using noise as a test pattern for determining image enhancement filters

Many digitally based medical imaging systems include both reconstruction algorithms and additional image filters designed to enhance certain image features. However, the manufacturers usually consider these algorithms and filters to be proprietory information. The purpose of this note is to describe a simple procedure for determining the spatial frequency response of these proprietary enhancement filters. The technique uses image noise as a test pattern. The procedure consists of acquiring a small number of noise-only data sets (say 10) of a uniform phantom and reconstructing the images using the different filters with repeated use of the noise data sets. A straightforward analysis then yields the enhancement filter frequency responses.     

基于同态滤波与直方图均衡化的超声图像增强

背景：同态滤波与直方图均衡化都能单独对图像进行增强处理,但各自还有许多不足之处。 目的：观察将同态滤波与直方图均衡化结合使用对增强图像的效果,减少单独使用这两种方法产生的缺点。 方法：在频域内,首先对超声图像分别进行高通同态滤波和低通同态滤波的分频处理;再将得到的两幅滤波图像线性组合;最后再对该组合图像进行全局的直方图均衡化得到最终的增强超声图像。 结果与结论：实验结果表明,超声图像经过上述方法的处理,边界十分突出明显,且整体视觉效果更明亮清晰。通过分析对比得到的均方根误差和信噪比数据,也证实了实验所用的超声图像增强法是十分切实有效的。     

Selective enhancement filters for nodules,vessels, and airway walls in two- and three-dimensional CT scans

Computer-aided diagnostic (CAD) schemes have been developed to assist radiologists in the early detection of lung cancer in radiographs and computed tomography (CT) images. In order to improve sensitivity for nodule detection, many researchers have employed a filter as a preprocessing step for enhancement of nodules. However, these filters enhance not only nodules, but also other anatomic structures such as ribs, blood vessels, and airway walls. Therefore, nodules are often detected together with a large number of false positives caused by these normal anatomic structures. In this study, we developed three selective enhancement filters for dot, line, and plane which can simultaneously enhance objects of a specific shape (for example, dot-like nodules) and suppress objects of other shapes (for example, line-like vessels). Therefore, as preprocessing steps, these filters would be useful for improving the sensitivity of nodule detection and for reducing the number of false positives. We applied our enhancement filters to synthesized images to demonstrate that they can selectively enhance a specific shape and suppress other shapes. We also applied our enhancement filters to real two-dimensional (2D) and three-dimensional (3D) CT images to show their effectiveness in the enhancement of specific objects in real medical images. We believe that the three enhancement filters developed in this study would be useful in the computerized detection of cancer in 2D and 3D medical images.     

基于自适应标记分水岭算法的肝脏CT图像自动分割

目的为减少人工交互提出了基于自适应标记分水岭的CT系列图像肝脏区域自动分割算法。方法首先对图像进行形态学重构运算以平滑图像,然后计算多尺度形态学梯度,同时提出利用梯度图像非零的局部极小值点的均值进行自适应标记提取,以避免分水岭的过分割和欠分割,再结合肝脏为最大的实质性脏器和相邻图像的相似性实现CT系列图像的肝区自动分割。结果该算法能自动、快速地提取CT系列图像中的肝脏区域。结论分水岭算法能准确定位区域的边缘,通过选择合适的阈值对梯度图像进行标记以抑制分水岭的过分割,实现医学图像中感兴趣区域的自动分割。     

Image filtering for improved dose resolution in CT polymer gel dosimetry

X-ray computed tomography (CT) has been established as a feasible method of performing dosimetry using polyacrylamide gels (PAGs). A small density change occurs in PAG upon irradiation that provides contrast in PAG CT images. However, low dose resolution limits the clinical usefulness of the technique. This work investigates the potential of using image filtering techniques on PAG CT images in order to reduce image noise and improve dose resolution. CT image noise for the scanner and protocol used for the gel images is analyzed and found to be Gaussian distributed and independent of the contrast level in the images. As a result, several filters for reducing spatially invariant noise are investigated: mean, median, midpoint, adaptive mean, alpha-trimmed mean, sigma mean, and a relatively new filter called SUSAN (smallest univalue segment assimilating nucleus). All filters are applied, using 3x3, 5x5, and 7x7 pixel masks, to a CT image of a PAG irradiated with a stereotactic radiosurgery dose distribution. The dose resolution within 95% confidence (D(delta)95%) is calculated and compared for each filtered image, as well the unfiltered image. In addition, the ability of the filters to maintain the spatial integrity of the dose distribution is evaluated and compared. Results clearly indicate that the filters are not equal in their ability to improve D(delta)95% or in their effect on the spatial integrity of the dose distribution. In general, increasing mask size improves D(delta)95% but simultaneously degrades spatial dose information. The mean filter provides the greatest improvement in D(delta)95%, but also the greatest loss of spatial dose information. The SUSAN, mean adaptive, and alpha-trimmed mean filters all provide comparable, but slightly poorer dose resolution. In addition, the SUSAN and adaptive filters both excel at maintaining the spatial distribution of dose and overall are the best performing filters for this application. The midpoint filter, normally useful for Gaussian noise, is poor all-round, dramatically distorting the dose distribution for masks greater than 3x3. The median filter, a common edge preserving noise reduction filter, performs moderately well, but artificially increases high dose gradients. The sigma filter preserves the spatial distribution of dose very well but is least effective at improving dose resolution. In summary, dose resolution can be significantly improved in CT PAG dosimetry through postprocessing of CT images using spatial noise reduction filters. However, such filters are not equal in their ability to improve dose resolution or to maintain the spatial integrity of the dose distribution and an appropriate filter must be chosen depending on clinical demands of the application.     

Design and evaluation of an external filter technique for exposure equalization in mammography.

We are developing an external filter method for equalizing x-ray exposure in the peripheral region of the breast. This method requires the use of only a limited number of custom-built filters for different breast shapes in a given view. This paper describes the design methodology for these external filters. The filter effectiveness was evaluated through a simulation study on 171 mediolateral and 196 craniocaudal view digitized mammograms and through imaging of a breast phantom. The degree of match between the simulated filter and the individual 3-D exposure profiles at the breast periphery was quantified. An analysis was performed to investigate the effect of filter misalignment. The simulation study indicates that the filter is effective in equalizing exposures for more than 80% of the breast images in our database. The tolerance in filter misalignment was estimated to be about +/- 2 mm for the CC view and +/- 1 mm for the MLO view at the image plane. Some misalignment artifacts were demonstrated with simulated filtered mammograms.     

Filters and Fourier analysis of gated blood pool studies: a search for the optimal combination

Fourier analysis of gated blood pool studies is performed after filtering the raw data by a spatial median 3 x 3, 9 x 9 or temporo-spatial 9 x 9 x 9 filter. 20 patients and a dynamic cardiac phantom were studied to determine the quantitative effects of these filters and of multiharmonic Fourier filtering (MHFF). The filtered MHFF data, with or without preprocessing, were compared with a 3 D or 2 D filter to the raw data using a chi 2 distribution. The MHFF (two or three harmonics) procedure applied to the raw data of patients without any preprocessing produced the smallest chi 2 value, thus demonstrating the very close relationship between filtered images and raw data. Preprocessing the raw data by the median filter also preserved the signal when two or three harmonics were applied, whereas the 3 D and 2 D (9 x 9) filters did not. The phantom study also demonstrated that MHFF preserved the signal better than any other preprocessing. The median filter introduced a smaller distortion than the 2 D (9 x 9) and 3 D filters. It is concluded that MHFF applied with two or three harmonics on the raw data or after preprocessing by a median (3 x 3) filter is the most successful way of preserving the real signal. It is believed that the other filters should be avoided. The clinical advantage of MHFF processing is to provide both very accurate filtering and parametric images.     

一种基于脉冲耦合神经网络的脉冲噪声滤波器设计

根据脉冲噪声与其邻域中图像灰度之间的明显差异性 ,本文提出了一种新的神经网络脉冲噪声滤波器设计方案。这种脉冲耦合神经网络PCNN滤波器比现有的PCNN逐次降噪方案迭代运算次数少 ,执行速度快。并与中值滤波器、全方位结构元约束层叠滤波器、全方位结构元形态闭 开最小、开 闭最大滤波器等现有的非线性滤波器进行实验比较 ,证明 ,该方案有更好的降噪性能 ,更重要的是比这些方案更有效地保持了图像的高频细节信息。     

Noise and signal decoupling in maximum-likelihood reconstructions and Metz filters for PET brain images

Images reconstructed with the maximum-likelihood-by-expectation-maximization (ML) algorithm have lower noise in some regions, particularly low count areas, compared with images reconstructed with filtered backprojection (FBP). The use of statistically correct noise model coupled with the positivity constraint in the ML algorithm provides this noise improvement, but whether this model confers a general advantage for ML over FBP with no noise model and any reconstruction filter, is unclear. We have studied the quantitative impact of the correct noise model in the ML algorithm applied to simulated and real PET fluoro-deoxyglucose (FDG) brain images, given a simplified but accurate reconstruction model with spatially invariant resolution. For FBP reconstruction, several Metz filters were chosen and images with different resolution were obtained depending on the order (1-400) of the Metz filters. Comparisons were made based on the mean Fourier spectra of the projection amplitudes, the noise-power spectra, and the mean region-of-interest signal and noise behaviour in the images. For images with resolution recovery beyond the intrinsic detector resolution, the noise increased significantly for FBP compared with ML. This indicates that in the process of signal recovery using ML, the noise is decoupled from the signal. Such noise decoupling is not possible for FBP. However, for image resolution equivalent to or less than the intrinsic detector resolution, FBP with Metz filters of various orders can achieve a performance similar to ML. The significance of the noise decoupling advantage in ML is dependent on the reconstructed image resolution required for specific imaging tasks.     

Selection of biorthogonal filters for image compression of MR images using wavelet packets

We present an analysis of different filter banks for the compression of magnetic resonance (MR) images of the human brain using wavelet packets based on biorthogonal filters. Initially, peak signal to noise ratio (PSNR) and normalized root mean square (RMS) error criteria are calculated for a series of images compressed with a 33:1 ratio, using filter banks based on biorthogonal wavelet packets. The results lead us to choose a few of these filter banks as optimal for image compression. One of these filters is employed to compress several images at four different compression ratios: 12.5:1, 25:1, 37.5:1 and 50:1. The quality of these images was evaluated by visual analysis by a group of seven experts who graded image quality on a 0-7 scale. Results show that using these filters, we can compress images to a rate of around 30:1 without introducing noticeable differences. Other applications for these filters are currently under study and include the compression/fusion of MR image stacks in order to obtain even better reductions in the amount of data needed to reconstruct complete MRI studies.     

Nonlinear filters applied on computerized axial tomography: theory and phantom images.

New nonlinear image processing techniques, in particular smoothing based on the understanding of the image, may create computerized tomography (CT) images of good quality using less radiation. Such techniques may be applied before the reconstruction and particularly after it. Current CT scanners use strong linear low-pass filters applied to the CT projections, reducing noise but also deteriorating the resolution of the image. The method in this study was to apply a weak low-pass filter on the projections, to perform the reconstruction, and only then to apply a nonlinear filter on the image. Various kinds of nonlinear filters were investigated based on the fact that the image is approximately piecewise constant. The filters were applied with many values of several parameters and the effects on the spatial resolution and the noise reduction were evaluated. The signal-to-noise ratio of a high-contrast phantom image processed were compared with the nonlinear filter, with the SNR of the phantom images obtained with the built-in CT linear filters in two scanning modes, the normal and the ultra high resolution modes. It was found that the nonlinear filters improve the SNR of the image, compared to the built-in filters, about three times for the normal mode and twice for the UHR scanning mode. The most successful filter on low-contrast phantom image was applied and it also seems to lead to promising results. These results seem to show that applying nonlinear filters on CT images might lead to better image quality than using the current linear filters.(ABSTRACT TRUNCATED AT 250 WORDS)     

Quantitative analysis of geometric-pattern features of interstitial infiltrates in digital chest radiographs: Preliminary results

We are developing a computerized method for detection and characterization of interstitial diseases based on a quantitative analysis of geometric features of various infiltrate patterns in digital chest radiographs. In our approach, regions of interest (ROIs) with 128 × 128 matrix size (22.4 mm × 22.4 mm) are automatically selected, covering peripheral lung regions. Next, nodular and linear opacities, which are the basic components of interstitial infiltrates, are identified from two processed images obtained by use of a multiple-level thresholding technique and a line enhancement filter, respectively. Finally, the total area of nodular opacities and the total length of linear opacities in each ROI are determined as measures of geometric pattern features. We have applied this computer analysis to 72 ROIs with normal and abnormal patterns that were classified in advance by six chest radiologists. Preliminary results indicate that the distribution of measures of geometric-pattern features correlate well with radiologists’ classification. These early results are encouraging, and further evaluation hopes to establish that this computerized method might prove useful to radiologists in their assessment of interstitial diseases.     

Displaying radiologic images on pesonal computers: Image processing and analysis

This is the fourth article of our series for radiologists and imaging scientists on displaying, manipulating, and analyzing radiologic images on personal computers. Classic image processing is divided into point, area, frame, and geometric processes. Point processes change image pixel values based on the value of the pixel of interest. Histogram equalization adjusts the pixel values in the image based on the distribution of pixel values. Area processes change the pixel of interest based on the values of the surrounding pixels, known as the neighborhood. Area processes using a convolution kernel are often used as image filters. Common convolution kernels include low-frequency, high-frequency, and edge-enhancement filters. Edge enhancement can be performed with convolution kernels such as shift and difference, gradient-directional and Laplacian filters, or with nonlinear methods such as Sobel's algorithm. Frame processes mathematically combine two or more images, often for noise reduction and background subtraction. Geometric processes alter the location of pixels within the image, but usually not the pixel values. Common radiologic applications of image processing include window width and window level adjustments (point process), adaptive histogram equalization (area process), unsharp masking (area process), computed radiography image processing (combined area and point processes), digital subtraction angiography (frame and geometric processes), region of interest analysis (area process), and image rotation (geometric process). As digital imaging becomes more widespread, radiologists need to understand the image processing that is fundamental to these modalities.     

Gray-scale edge detection for gastric tumor pathologic cell images by morphological analysis

For the purpose of analyzing gastric tumor pathologic cell images, a novel method is developed with gray-scale edge detection of mathematical morphology in this study. In combination with texture features of the image under investigation, this paper works on edge detection with various structuring elements (SEs) and gray-scale values. The results of the experiment are presented, and we found several advantages by using the morphological edge detection scheme for the analysis of gastric tumor pathologic cell images. Meanwhile, the results of the binary morphological edge detection are given for comparison.     

A computer-aided temporal and dynamic subtraction technique of the liver for detection of small hepatocellular carcinomas on abdominal CT images

It is often difficult for radiologists to identify small hepatocellular carcinomas (HCCs) due to insufficient contrast enhancement. Therefore, we have developed a new computer-aided temporal and dynamic subtraction technique to enhance small HCCs, after automatically selecting images set at the same anatomical position from the present (non-enhanced and arterial-phase CT images) and previous images. The present study was performed with CT images from 14 subjects. First, we used template-matching based on similarities in liver shape between the present (non-enhanced and arterial-phase CT images) and previous arterial-phase CT images at the same position. Temporal subtraction images were then obtained by subtraction of the previous image from the present image taken at the same position of the liver. Dynamic subtraction images were also obtained by subtraction of non-enhanced CT images from arterial-phase CT images taken at the same position of the liver. Twenty-one of 22 nodules (95.5%) with contrast enhancement were visualized in temporal and dynamic subtraction images. Compared with present arterial-phase CT images, increases of 150% and 140% in nodule-to-liver contrast were observed on dynamic and temporal subtraction images, respectively. These subtraction images may be useful as reference images in the detection of small moderately differentiated HCCs.     

A Wiener filter for nuclear medicine images

To improve the quality of digital nuclear medicine images, we have developed a new implementation of the Wiener restoration filter. The Wiener filter uses as its optimality criterion the minimization of the mean-square error between the undistorted image of the object and the filtered image. In order to form this filter, the object and noise power spectrums are needed. The noise power spectrum for the count-dependent Poisson noise of nuclear medicine images is shown to have a constant average magnitude equal to the total count in the image. The object power spectrum is taken to be the image power spectrum minus the total count, except in the noise dominated region of the image power spectrum where a least-squares-fitted exponential is used. Processing time is kept to a clinically acceptable time frame through use of an array processor. Pronounced noise suppression and detail enhancement are noted with use of this filter with clinical images.     

The derivation and verification of a non-stationary, optimal smoothing filter for nuclear medicine image data

A non-stationary optimal smoothing filter for digital nuclear medicine image data, degraded by Poisson noise, has been derived and applied to temporal simulated and clinical gated blood pool study (GBPS) data. The derived filter is automatically calculated from a large group (library) of similar GBPS which are representative of all studies acquired according to the same protocol in a defined patient population (the ensemble). The filter is designed to minimize the mean-square difference between the filtered data and the true image values; it provides an optimal trade-off between noise reduction and signal degradation for members of the ensemble. The filter is evaluated using a computer simulated ensemble of GBPS. Libraries of Poisson-degraded and non-degraded studies were generated. Libraries of up to 400 Poisson-degraded simulated studies were used to estimate optimal temporal filters that, when applied to Poisson-degraded members of the ensemble not included in the libraries, reduced the mean-square error in the raw data by 65%. When the non-degraded studies were used instead to compute the optimal filter values, the corresponding reduction in the error was 83%. Libraries of previously acquired clinical GBPS were then used to estimate optimal temporal filters for an ensemble of similarly acquired studies. These filters were subsequently applied to studies of 13 patients (not in the original libraries) who received multiple sequential repeat studies. Comparisons of both the filtered and raw data to averages of the repeat studies demonstrated that optimal filters calculated from 400 and 800 clinical studies reduced the mean-square error in the clinical data by 56% and 63% respectively.     

用于脂肪肝分级诊断的超声图像增强算法

提出一种针对脂肪肝分级诊断的超声图像增强算法,提高超声图像对脂肪肝的分级诊断能力。通过分析正常肝和不同程度脂肪肝在超声图像上的表现特点,本研究提出了一种将灰度的分段线性变换方法和改进的局部对比度增强方法相结合的图像增强方法,该方法首先通过灰度分段线性变换处理来凸显图像的脂肪肝特征,然后利用改进的局部对比度增强处理来增强图像的细节,提高图像的整体质量。应用该方法分别对正常肝和轻度、中度以及重度脂肪肝超声图像进行增强处理并对比原图像,结果表明,增强后图像的脂肪肝特征更加凸显,图像细节得到了提高,不同程度脂肪肝图像之间的差异更明显,具有更好的脂肪肝诊断信息。本研究提出的方法可提高超声图像分级诊断脂肪肝的能力,具有一定的实际应用价值。     

Enhancing obstetric and gynecology ultrasound images by adaptation of the speckle reducing anisotropic diffusion filter

OBJECTIVE: So far there is no ideal speckle reduction filtering technique that is capable of enhancing and reducing the level of noise in medical ultrasound (US) images, while efficiently responding to medical experts' validation criteria which quite often include a subjective component. This paper presents an interactive tool called evolutionary speckle reducing anisotropic diffusion filter (EVOSRAD) that performs adaptive speckle filtering on ultrasound B-mode still images. The medical expert runs the algorithm interactively, having a permanent control over the output, and guiding the filtering process towards obtaining enhanced images that agree to his/her subjective quality criteria. METHODS AND MATERIAL: We employ an interactive evolutionary algorithm (IGA) to adapt on-line the parameters of a speckle reducing anisotropic diffusion (SRAD) filter. For a given input US image, the algorithm evolves the parameters of the SRAD filter according to subjective criteria of the medical expert who runs the interactive algorithm. The method and its validation are applied to a test bed comprising both real and simulated obstetrics and gynecology (OB/GYN) ultrasound images. RESULTS: The potential of the method is analyzed in comparison to other speckle reduction filters: the original SRAD filter, the anisotropic diffusion, offset and median filters. Results obtained show the good potential of the method on several classes of OB/GYN ultrasound images, as well as on a synthetic image simulating a real fetal US image. Quality criteria for the evaluation and validation of the method include subjective scoring given by the medical expert who runs the interactive method, as well as objective global and local quality criteria. CONCLUSIONS: The method presented allows the medical expert to design its own filters according to the degree of medical expertise as well as to particular and often subjective assessment criteria. A filter is designed for a given class of ultrasound images and for a given medical expert who will later use the respective filter in clinical practice. The process of designing a filter is simple and employs an interactive visualization and scoring stage that does not require image processing knowledge. Results show that filters tailored using the presented method achieve better quality scores than other more generic speckle filtering techniques.