Wavelet-based statistical approach for speckle reduction in medical ultrasound images

A novel speckle-reduction method is introduced, based on soft thresholding of the wavelet coefficients of a logarithmically transformed medical ultrasound image. The method is based on the generalised Gaussian distributed (GGD) modelling of sub-band coefficients. The method used was a variant of the recently published BayesShrink method by Chang and Vetterli, derived in the Bayesian framework for denoising natural images. It was scale adaptive, because the parameters required for estimating the threshold depend on scale and sub-band data. The threshold was computed by Kσ/σ_x, where σ and σ_x were the standard deviation of the noise and the sub-band data of the noise-free image, respectively, and K was a scale parameter. Experimental results showed that the proposed method outperformed the median filter and the homomorphic Wiener filter by 29% in terms of the coefficient of correlation and 4% in terms of the edge preservation parameter. The numerical values of these quantitative parameters indicated the good feature preservation performance of the algorithm, as desired for better diagnosis in medical image processing.     

Monte Carlo simulations of clinical PET and SPECT scans: impact of the input data on the simulated images

Monte Carlo simulations of emission tomography have proven useful to assist detector design and optimize acquisition and processing protocols. The more realistic the simulations, the more straightforward the extrapolation of conclusions to clinical situations. In emission tomography, accurate numerical models of tomographs have been described and well validated under specific operating conditions (collimator, radionuclide, acquisition parameters, count rates, etc). When using these models under these operating conditions, the realism of simulations mostly depends on the activity distribution used as an input for the simulations. It has been proposed to derive the input activity distribution directly from reconstructed clinical images, so as to properly model the heterogeneity of the activity distribution between and within organs. However, reconstructed patient images include noise and have limited spatial resolution. In this study, we analyse the properties of the simulated images as a function of the properties of the reconstructed images used to define the input activity distributions in (18)F-FDG PET and (131)I SPECT simulations. The propagation through the simulation/reconstruction process of the noise and spatial resolution in the input activity distribution was studied using simulations. We found that the noise properties of the images reconstructed from the simulated data were almost independent of the noise in the input activity distribution. The spatial resolution in the images reconstructed from the simulations was slightly poorer than that in the input activity distribution. However, using high-noise but high-resolution patient images as an input activity distribution yielded reconstructed images that could not be distinguished from clinical images. These findings were confirmed by simulated highly realistic (131)I SPECT and (18)F-FDG PET images from patient data. In conclusion, we demonstrated that (131)I SPECT and (18)F-FDG PET images indistinguishable from real scans can be simulated using activity maps with spatial resolution higher than that used in routine clinical applications.     

A comparison of MR-based attenuation correction in PET versus SPECT

Attenuation correction (AC) is a critical step in the reconstruction of quantitatively accurate positron emission tomography (PET) and single photon emission computed tomography (SPECT) images. Several groups have proposed magnetic resonance (MR)-based AC algorithms for application in hybrid PET/MR systems. However, none of these approaches have been tested on SPECT data. Since SPECT/MR systems are under active development, it is important to ascertain whether MR-based AC algorithms validated for PET can be applied to SPECT. To investigate this issue, two imaging experiments were performed: one with an anthropomorphic chest phantom and one with two groups of canines. Both groups of canines were imaged from neck to abdomen, one with PET/CT and MR (n = 4) and the other with SPECT/CT and MR (n = 4), while the phantom was imaged with all modalities. The quality of the nuclear medicine reconstructions using MR-based attenuation maps was compared between PET and SPECT on global and local scales. In addition, the sensitivity of these reconstructions to variations in the attenuation map was ascertained. On both scales, it was found that the SPECT reconstructions were of higher fidelity than the PET reconstructions. Further, they were less sensitive to changes to the MR-based attenuation map. Thus, MR-based AC algorithms that have been designed for PET/MR can be expected to demonstrate improved performance when used for SPECT/MR.     

A bootstrap approach for assessing the uncertainty of outcome probabilities when using a scoring system

Background  

Scoring systems are a very attractive family of clinical predictive models, because the patient score can be calculated without using any data processing system. Their weakness lies in the difficulty of associating a reliable prognostic probability with each score. In this study a bootstrap approach for estimating confidence intervals of outcome probabilities is described and applied to design and optimize the performance of a scoring system for morbidity in intensive care units after heart surgery.     

Porous phantoms for PET and SPECT performance evaluation and quality assurance

Characterization of PET and SPECT imaging performance often requires phantoms with complex radionuclide distributions. For example, lesion detection studies use multiple spherical regions of specific target-to-background ratios to simulate cancerous lesions. Such complex distributions are typically created using phantoms with multiple fillable chambers. However, such phantoms are typically difficult and time-consuming to prepare accurately and reproducibly. A new approach using a single-chamber phantom with a porous core can overcome these difficulties. METHODS: Prototypes of two designs of porous core phantoms were produced and evaluated. The "hot spheres" phantom contained a multitude of simulated spherical lesions with diameters ranging from 6.35 to 25.4 mm ("multi-resolution" slice) and with lesion-to-background ratios ranging from 1.6 to 4.4 ("multi-contrast" slice). The "multi-attenuation" phantom consisted of two halves. One half contained a porous core to produce regions of different attenuation but uniform activity. The other half mimicked the NEMA-94 design with cold inserts of different attenuation. RESULTS: Both phantoms produced the expected radionuclide distributions while requiring the preparation of only a single radionuclide solution and with much reduced preparation time. In images taken on clinical PET and SPECT scanners, the porous core structures were found to contribute negligible background noise or artifact. The measured lesion-to-background ratios from the hot spheres phantom differed slightly from calculated values, with the differences attributed mainly to uncertainty in pore diameter. The measured attenuation coefficients from the multi-attenuation phantom agreed well with expected values. However, it was found that trapped air bubbles due to manufacturing defects in the porous core could potentially cause quantitative errors. CONCLUSION: The hot spheres and multi-attenuation porous phantoms exhibited a wide range of imaging features providing thorough tests of lesion detection and of attenuation and scatter correction accuracy. Because the local activity concentration is set by the relative volume of radionuclide solution in the porous core, the quantitative accuracy is limited mainly by mechanical tolerance, and strict quality control during manufacturing is essential. Nonetheless, the single-chamber design of the porous core phantoms is inherently more reproducible and more practical for routine use compared to conventional multi-chamber phantoms.     

A methodology for generating normal and pathological brain perfusion SPECT images for evaluation of MRI/SPECT fusion methods: application in epilepsy

Quantitative evaluation of brain MRI/SPECT fusion methods for normal and in particular pathological datasets is difficult, due to the frequent lack of relevant ground truth. We propose a methodology to generate MRI and SPECT datasets dedicated to the evaluation of MRI/SPECT fusion methods and illustrate the method when dealing with ictal SPECT. The method consists in generating normal or pathological SPECT data perfectly aligned with a high-resolution 3D T1-weighted MRI using realistic Monte Carlo simulations that closely reproduce the response of a SPECT imaging system. Anatomical input data for the SPECT simulations are obtained from this 3D T1-weighted MRI, while functional input data result from an inter-individual analysis of anatomically standardized SPECT data. The method makes it possible to control the 'brain perfusion' function by proposing a theoretical model of brain perfusion from measurements performed on real SPECT images. Our method provides an absolute gold standard for assessing MRI/SPECT registration method accuracy since, by construction, the SPECT data are perfectly registered with the MRI data. The proposed methodology has been applied to create a theoretical model of normal brain perfusion and ictal brain perfusion characteristic of mesial temporal lobe epilepsy. To approach realistic and unbiased perfusion models, real SPECT data were corrected for uniform attenuation, scatter and partial volume effect. An anatomic standardization was used to account for anatomic variability between subjects. Realistic simulations of normal and ictal SPECT deduced from these perfusion models are presented. The comparison of real and simulated SPECT images showed relative differences in regional activity concentration of less than 20% in most anatomical structures, for both normal and ictal data, suggesting realistic models of perfusion distributions for evaluation purposes. Inter-hemispheric asymmetry coefficients measured on simulated data were found within the range of asymmetry coefficients measured on corresponding real data. The features of the proposed approach are compared with those of other methods previously described to obtain datasets appropriate for the assessment of fusion methods.     

Testing of the assisting software for radiologists analysing head CT images: lessons learned

Background

Assessing a plan for user testing and evaluation of the assisting software developed for radiologists.

Methods

Test plan was assessed in experimental testing, where users performed reporting on head computed tomography studies with the aid of the software developed. The user testing included usability tests, questionnaires, and interviews. In addition, search relevance was assessed on the basis of user opinions.

Results

The testing demonstrated weaknesses in the initial plan and enabled improvements. Results showed that the software has acceptable usability level but some minor fixes are needed before larger-scale pilot testing. The research also proved that it is possible even for radiologists with under a year’s experience to perform reporting of non-obvious cases when assisted by the software developed. Due to the small number of test users, it was impossible to assess effects on diagnosis quality.

Conclusions

The results of the tests performed showed that the test plan designed is useful, and answers to the key research questions should be forthcoming after testing with more radiologists. The preliminary testing revealed opportunities to improve test plan and flow, thereby illustrating that arranging preliminary test sessions prior to any complex scenarios is beneficial.

     

A comparative study of surface- and volume-based techniques for the automatic registration between CT and SPECT brain images

Image registration of multimodality images is an essential task in numerous applications in three-dimensional medical image processing. Medical diagnosis can benefit from the complementary information in different modality images. Surface-based registration techniques, while still widely used, were succeeded by volume-based registration algorithms that appear to be theoretically advantageous in terms of reliability and accuracy. Several applications of such algorithms for the registration of CT-MRI, CT-PET, MRI-PET, and SPECT-MRI images have emerged in the literature, using local optimization techniques for the matching of images. Our purpose in this work is the development of automatic techniques for the registration of real CT and SPECT images, based on either surface- or volume-based algorithms. Optimization is achieved using genetic algorithms that are known for their robustness. The two techniques are compared against a well-established method, the Iterative Closest Point-ICP. The correlation coefficient was employed as an independent measure of spatial match, to produce unbiased results. The repeated measures ANOVA indicates the significant impact of the choice of registration method on the magnitude of the correlation (F = 4.968, p = 0.0396). The volume-based method achieves an average correlation coefficient value of 0.454 with a standard deviation of 0.0395, as opposed to an average of 0.380 with a standard deviation of 0.0603 achieved by the surface-based method and an average of 0.396 with a standard deviation equal to 0.0353 achieved by ICP. The volume-based technique performs significantly better compared to both ICP (p<0.05, Neuman Keuls test) and the surface-based technique (p<0.05, Neuman-Keuls test). Surface-based registration and ICP do not differ significantly in performance.     

人脑MRI和PET图像的融合方法

目的：研究MRI和PET医学图像融合的方法并且应用在人脑神经解剖中。方法：以Chamfer matching为基本配准方法和以模糊数学为理论基础的融合方法。结果：使得融合后图像有很强的抗配准偏差能力,并且能极大程度地保留原来解剖性信息图像和功能性图像的信息。能根据融合算子的不同组合,得到倾向于不同检查仪器结果的图像。结论：通过本文方法得出的融合后图像,能在人脑解剖结构上显示出该区域的功能状态,对于了解该结构的存在、变异、病变提供了很大的便利。对人脑神经解剖学研究和临床应用都有很大的理论和实践意义。     

Diagnostic value of SPECT, PET and PET/CT in the diagnosis of coronary artery disease: A systematic review

Purpose:

The purpose of the study was to investigate the diagnostic value of SPECT, PET and PET/CT in the diagnosis of coronary artery disease, based on a systematic review.

Material and Methods:

A search of PubMed/Medline and Sciencedirect databases in the English-language literature published over the last 24 years was performed. Only studies with at least 10 patients comparing SPECT, PET or combined PET/CT with invasive coronary angiography in the diagnosis of coronary artery disease (50% stenosis) were included for analysis. Sensitivities and specificities estimates pooled across studies were analysed using a Chi-square test.

Results:

Twenty-five studies met the selection criteria and were included for the analysis. Ten studies were performed with SPECT alone; while another six studies were performed with PET alone. Five studies were carried out with both PET and SPECT modalities, and the remaining four studies were investigated with integrated PET-CT. The mean value of sensitivity, specificity and accuracy of these imaging modalities for the diagnosis of coronary artery disease was 82% (95%CI: 76 to 88), 76% (95%CI: 70 to 82) and 83% (95%CI: 77 to 89) for SPECT; 91% (95%CI: 85 to 97), 89% (95%CI: 83 to 95) and 89% (95%CI: 83 to 95) for PET; and 85% (95%CI: 79 to 90), 83% (95%CI: 77 to 89) and 88% (95%CI: 82 to 94) for PET/CT, respectively. The diagnostic accuracy of these imaging modalities was dependent on the radiotracers used in these studies, with ammonia resulting in the highest diagnostic value.

Conclusion:

Our review shows that PET has high diagnostic value for diagnosing coronary artery disease, and this indicates that it is a valuable technique for both detection and prediction of coronary artery disease.     

A hybrid strategy to integrate surface-based and mutual-information-based methods for co-registering brain SPECT and MR images

Co-registration of brain SPECT and MR images has been used extensively in clinical applications. The complementary features of two major co-registration methods—surface- and mutual-information-based (MI-based)—motivated us to study a hybrid-based scheme that uses the surface-based method to achieve a quick alignment, followed by the MI-based method for fine tuning. Computer simulations were conducted to evaluate the accuracy and robustness of surface-, MI-, and hybrid-based registration methods by designing different levels of noise and mismatch in the registration experiments. Results demonstrated that the hybrid surface-MI-based scheme outperforms both the surface- and MI-based methods in providing superior accuracy and success rates. Specifically, the translational and rotational errors were no more than 1 mm and 2°, respectively, with consistent success rates over 98%. Besides, the hybrid-based method saved 12–53% of the computation efforts, compared with using the MI-based method alone. We recommend the use of hybrid-based method when the orientational differences between the floating and reference images exceed 10°.     

A quantitative approach for analysing bone modelling patterns from craniofacial surfaces in hominins

Bone size and shape arise throughout ontogeny as a result of the coordinated activity of osteoblasts and osteoclasts, responsible for bone deposition and resorption, and growth displacements. The modelling processes leave specific microstructural features on the bone surface, which can be used to infer the mechanisms shaping craniofacial traits in extinct and extant species. However, the analysis of bone surfaces from fossils and archaeological samples faces some difficulties related to the bone loss caused by taphonomic factors, and the lack of formal methods for estimating missing information and comparing the patterns of bone modelling among several specimens and samples. The present study provides a new approach for the quantitative analysis of bone formation and resorption patterns obtained from craniofacial surfaces. First, interpolation techniques were used to estimate missing data on high‐resolution replicas of the left maxilla in a sample of sub‐adult and adult modern humans and sub‐adult fossil hominins. The performance of this approach was assessed by simulating variable amounts of missing data. Then, we applied measures of dispersion and central tendency to represent the variation and average pattern of bone modelling within samples. The spatial interpolation resulted in reliable estimations of the type of cell activity (deposition or resorption) in the missing areas, even when large extensions of the bone surface were lost. The quantification of the histological data allowed us to integrate the information of different specimens and depict the areas with higher and lower variation in the bone modelling pattern of the maxilla among specimens. Overall, the main advantages of the quantitative approach used here for generating bone modelling patterns are the high replicability and the possibility of incorporating variation among specimens into the comparisons among samples.     

A hybrid tissue segmentation approach for brain MR images

A novel hybrid algorithm for the tissue segmentation of brain magnetic resonance images is proposed. The core of the algorithm is a probabilistic neural network (PNN) in which weighting factors are added to the summation layer, such that partial volume effects can be taken into account in the modeling process. The mean vectors for the probability density function estimation and the corresponding weighting factors are generated by a hierarchical scheme involving a self-organizing map neural network and an expectation maximization algorithm. Unlike conventional PNN, this approach circumvents the need for training sets. Tissue segmentation results from various algorithms are compared and the effectiveness and robustness of the proposed approach are demonstrated.     

A statistical approach to the histopathologic diagnosis of synovitis

Synovial tissue from 393 operations on diseased joints were scored for each of 37 histopathologic variables; the patients fell into 10 clinically defined diagnostic categories. The five most populous of these diagnostic groups provided a data set for calculating diagnoses based on histopathologic findings. For each pair of clinical diagnostic categories we derived a linear discriminant function dependent solely on the histopathologic assessments. The functions gave a numerical score for segregating each case into one of the diagnostic categories. The histopathology based diagnoses agreed with the clinical diagnoses in 54 to 78 per cent of the cases. Illustrations of the application of this technique are presented.     

A statistical approach to the discrimination of the cranial form

In a study of lateral cephalographic dimensions, their multivariate statistical analysis was shown to discriminate between patients with gross cranial anomalies and controls exhibiting no abnormal morphology. Although the degree of discrimination depended upon the group of dimensions included in the analysis, discrimination was more consistent than when compared using univariate statistical techniques. As cephalographic dimensions combine both size and shape parameters together, however, the interpretation of such contrasts must await more specific analytic techniques development.     

Scatter-to-primary based scatter fractions for transmission-dependent convolution subtraction of SPECT images

In single photon emission computed tomography (SPECT), transmission-dependent convolution subtraction has been shown to be useful when correcting for scattered events. The method is based on convolution subtraction, but includes a matrix of scatter fractions instead of a global scatter fraction. The method can be extended to iteratively improve the scatter estimate, but in this note we show that this requires a modification of the theory to use scatter-to-total scatter fractions for the first iteration only and scatter-to-primary fractions thereafter. To demonstrate this, scatter correction is performed on a Monte Carlo simulated image of a point source of activity in water. The modification of the theory is compared to corrections where the scatter fractions are based on the scatter-to-total ratio, using one and ten iterations. The resulting ratios of subtracted to original counts are compared to the true scatter-to-total ratio of the simulation and the most accurate result is found for our modification of the theory.     

Stereotactic imaging for radiotherapy: accuracy of CT,MRI, PET and SPECT

CT, MRI, PET and SPECT provide complementary information for treatment planning in stereotactic radiotherapy. Stereotactic correlation of these images requires commissioning tests to confirm the localization accuracy of each modality. A phantom was developed to measure the accuracy of stereotactic localization for CT, MRI, PET and SPECT in the head and neck region. To this end. the stereotactically measured coordinates of structures within the phantom were compared with their mechanically defined coordinates. For MRI, PET and SPECT, measurements were performed using two different devices. For MRI, T1- and T2-weighted imaging sequences were applied. For each measurement, the mean radial deviation in space between the stereotactically measured and mechanically defined position of target points was determined. For CT, the mean radial deviation was 0.4 +/- 0.2 mm. For MRI, the mean deviations ranged between 0.7 +/- 0.2 mm and 1.4 +/- 0.5 mm, depending on the MRI device and the imaging sequence. For PET, mean deviations of 1.1 +/- 0.5 mm and 2.4 +/- 0.3 mm were obtained. The mean deviations for SPECT were 1.6 +/- 0.5 mm and 2.0 +/- 0.6 mm. The phantom is well suited to determine the accuracy of stereotactic localization with CT, MRI, PET and SPECT in the head and neck region. The obtained accuracy is well below the physical resolution for CT, PET and SPECT, and of comparable magnitude for MRI. Since the localization accuracy may be device dependent, results obtained at one device cannot be generalized to others.     

A statistical approach for determination of time plane features from digitized ECG

This paper illustrates a method for time-plane feature extraction from digitized ECG sample using statistical approach. The algorithm detects the position and magnitude of the QRS complex, P and T wave for a single lead ECG dataset. The processing is broadly based on relative comparison of magnitude and slopes of ECG samples. Then the baseline modulation in the dataset is removed. The R-peak detection and baseline modulation is tested MIT-BIH arrhythmia database as well as 12-lead datasets in MIT-PTB database (PTBDB) and available under Physionet. The overall accuracy obtained is more than 99%.