Ultrasonic tissue characterization via 2-D spectrum analysis: theory and in vitro measurements

Ultrasound wavelet spectra enable direct tissue recognition and full-color visualization

Traditional brightness-mode ultrasound imaging is primarily constrained by the low specificity among tissues and the inconsistency among sonographers. The major cause is the imaging method that represents the amplitude of echoes as brightness and ignores other detailed information, leaving sonographers to interpret based on organ contours that depend highly on specific imaging planes. Other ultrasound imaging modalities, color Doppler imaging or shear wave elastography, overlay motion or stiffness information to brightness-mode images. However, tissue-specific scattering properties and spectral patterns remain unknown in ultrasound imaging. Here we demonstrate that the distribution (size and average distance) of scattering particles leads to characteristic wavelet spectral patterns, which enables tissue recognition and high-contrast ultrasound imaging. Ultrasonic wavelet spectra from similar particle distributions tend to cluster in the eigenspace according to principal component analysis, whereas those with different distributions tend to be distinguishable from one another. For each distribution, a few wavelet spectra are unique and act as a fingerprint to recognize the corresponding tissue. Illumination of specific tissues and organs with designated colors according to the recognition results yields high-contrast ultrasound imaging. The fully-colorized tissue-specific ultrasound imaging potentially simplifies the interpretation and promotes consistency among sonographers, or even enables the applicability for non-professionals.

Attenuation Coefficient Parameter Computations for Tissue Composition Assessment of Carotid Atherosclerotic Plaque in Vivo

Quantitative ultrasound has been used to assess carotid plaque tissue composition. Here, we compute the attenuation coefficient (AC) in vivo with the optimum power spectral shift estimator (OPSSE) and reference phantom method (RPM), extract AC parameters and form parametric maps. Differences between OPSSE and RPM AC parameters are computed. Relationships between AC parameters, surgical scores and histopathology assessments are examined. Kendall's τ correlations between OPSSE AC and surgical scores are significant, including those between cholesterol and Standard Deviation (adjusted p = 0.038); thrombus and Minimum (adjusted p = 0.002), Maximum (adjusted p = 0.021) and Standard Deviation (adjusted p = 0.001); ulceration and Average (adjusted p = 0.033), Median (unadjusted p = 0.013), Maximum (unadjusted p = 0.039) and Mode (adjusted p = 0.009). The strongest correlations with histopathology are percentage cholesterol and Median OPSSE (unadjusted p = 0.007); percentage hemorrhage and Minimum OPSSE (adjusted p < 0.001); hemosiderin score and Median OPSSE (adjusted p = 0.010); and percentage calcium and Percentage Non-physical RPM Pixels (unadjusted p = 0.014). Kruskal-Wallis H and Dunn's post hoc tests have the ability to distinguish between groups (p < 0.05). Results suggest AC parameters may assist in vivo evaluation of carotid plaque vulnerability.

Tissue classification in intercostal and paravertebral ultrasound using spectral analysis of radiofrequency backscatter

Paravertebral and intercostal nerve blocks have experienced a resurgence in popularity. Ultrasound has become the gold standard for visualization of the needle during injection of the analgesic, but the intercostal artery and vein can be difficult to visualize. We investigated the use of spectral analysis of raw radiofrequency (RF) ultrasound signals for identification of the intercostal vessels and six other tissue types in the intercostal and paravertebral spaces. Features derived from the one-dimensional spectrum, two-dimensional spectrum, and cepstrum were used to train four different machine learning algorithms. In addition, the use of the average normalized spectrum as the feature set was compared with the derived feature set. Compared to a support vector machine (SVM) (74.2%), an artificial neural network (ANN) (68.2%), and multinomial analysis (64.1%), a random forest (84.9%) resulted in the most accurate classification. The accuracy using a random forest trained with the first 15 principal components of the average normalized spectrum was 87.0%. These results demonstrate that using a machine learning algorithm with spectral analysis of raw RF ultrasound signals has the potential to provide tissue characterization in intercostal and paravertebral ultrasound.

Influence of Ultrasound System and Gain on Grayscale Median Values

OBJECTIVES

The purpose of this study was to determine the reliability of grayscale median (GSM) measurements across different ultrasound (US) systems and effects of gain on GSM values.

METHODS

Two vessels in a grayscale vascular phantom were imaged with 7 US systems at 3 gain settings. Two human participants were imaged at 3 gain settings. Each image was normalized, standardized, and segmented by expert and novice readers using grayscale analysis software. The concordance correlation coefficient (CCC) assessed agreement of GSM values for each system across gain settings and vessels and between readers. The intraclass correlation coefficient (ICC) assessed system-level reader concordance across gain settings and vessels. A general linear mixed model for repeated measures was used to assess within- and between-system mean GSM values.

RESULTS

Grayscale median measurements performed on images from the same US system yielded excellent (CCC) (95% confidence intervals): 0.85 (0.75, 0.92) to 0.96 (0.92, 0.98). ICC per system were 0.94 to 0.98 for the expert reader and 0.85 to 0.95 for the novice reader. Gain adjustments above and below an optimal setting contributed to significantly different intrasystem GSM values on 4 of 7 systems in the near zone and 5 of 7 systems in the far zone (P < .05). Intersystem GSM values differed on 5 of 7 systems (P < .05). Images from the human participants showed differences in GSM values at optimum gain values ± 10 dB/%.

CONCLUSIONS

Grayscale median measurements are highly reproducible when obtained from the same US system with similar gain settings. Grayscale median values differ significantly across gain values and between systems. Researchers should consider the impact of US system and gain settings on GSM values when working to minimize system- and operator-dependent factors.

Spectral Analysis of Ultrasound Radiofrequency Backscatter for the Detection of Intercostal Blood Vessels

Spectral analysis of ultrasound radiofrequency backscatter has the potential to identify intercostal blood vessels during ultrasound-guided placement of paravertebral nerve blocks and intercostal nerve blocks. Autoregressive models were used for spectral estimation, and bandwidth, autoregressive order and region-of-interest size were evaluated. Eight spectral parameters were calculated and used to create random forests. An autoregressive order of 10, bandwidth of 6 dB and region-of-interest size of 1.0 mm resulted in the minimum out-of-bag error. An additional random forest, using these chosen values, was created from 70% of the data and evaluated independently from the remaining 30% of data. The random forest achieved a predictive accuracy of 92% and Youden's index of 0.85. These results suggest that spectral analysis of ultrasound radiofrequency backscatter has the potential to identify intercostal blood vessels. (jokling@siue.edu) © 2018 World Federation for Ultrasound in Medicine and Biology.

Automatic Differential Diagnosis of Melanocytic Skin Tumors Using Ultrasound Data

We describe a novel automatic diagnostic system based on quantitative analysis of ultrasound data for differential diagnosis of melanocytic skin tumors. The proposed method has been tested on 160 ultrasound data sets (80 of malignant melanoma and 80 of benign melanocytic nevi). Acoustical, textural and shape features have been evaluated for each segmented lesion. Using parameters selected according to Mahalanobis distance and linear support vector machine classifier, we are able to differentiate malignant melanoma from benign melanocytic skin tumors with 82.4% accuracy (sensitivity = 85.8%, specificity = 79.6%). The results indicate that high-frequency ultrasound has the potential to be used for differential diagnosis of melanocytic skin tumors and to provide supplementary information on lesion penetration depth. The proposed system can be used as an additional tool for clinical decision support to improve the early-stage detection of malignant melanoma.

A restoration framework for ultrasonic tissue characterization

Ultrasonic tissue characterization has become an area of intensive research. This procedure generally relies on the analysis of the unprocessed echo signal. Because the ultrasound echo is degraded by the non-ideal system point spread function, a deconvolution step could be employed to provide an estimate of the tissue response that could then be exploited for a more accurate characterization. In medical ultrasound, deconvolution is commonly used to increase diagnostic reliability of ultrasound images by improving their contrast and resolution. Most successful algorithms address deconvolution in a maximum a posteriori estimation framework; this typically leads to the solution of l(2)-norm or (1)-norm constrained optimization problems, depending on the choice of the prior distribution. Although these techniques are sufficient to obtain relevant image visual quality improvements, the obtained reflectivity estimates are, however, not appropriate for classification purposes. In this context, we introduce in this paper a maximum a posteriori deconvolution framework expressly derived to improve tissue characterization. The algorithm overcomes limitations associated with standard techniques by using a nonstandard prior model for the tissue response. We present an evaluation of the algorithm performance using both computer simulations and tissue-mimicking phantoms. These studies reveal increased accuracy in the characterization of media with different properties. A comparison with state-of-the-art Wiener and l(1)-norm deconvolution techniques attests to the superiority of the proposed algorithm.

A feasibility study of novel ultrasonic tissue characterization for prostate-cancer diagnosis: 2D spectrum analysis of in vivo data with histology as gold standard

This study demonstrates the feasibility of using a novel 2D spectrum ultrasonic tissue characterization (UTC) technique for prostate-cancer diagnosis. Normalized 2D spectra are computed by performing Fourier transforms along the range (beam) and the cross-range directions of the digital radio-frequency echo data, then dividing by a reference spectrum. This 2D spectrum method provides axial and lateral information of tissue microstructures, an improvement over the current 1D spectrum analysis which only provides axial information. A pilot study was conducted on four prostate-cancer patients who underwent radical prostatectomies. Cancerous and noncancerous regions of interest, identified through histology, were compared using four 2D spectral parameters: peak value and 3 dB width of the radially integrated spectral power (RISP), slope and intercept of the angularly integrated spectral power (AISP). For noncancerous and cancerous prostatic tissues, respectively, our investigation yielded 23 +/- 1 and 26 +/- 1 dB for peak value of RISP, 7.8 +/- 0.5 degrees and 7.6 +/- 0.6 degrees for 3 dB of RISP, -2.1 +/- 0.2 and -2.7 +/- 0.4 dB/MHz for slope of AISP, and 92 +/- 5 and 112 +/- 6 dB for intercept of AISP. Preliminary results indicated that 2D spectral UTC has the potential for identifying tumor-bearing regions within the prostate gland.

Tendinopathy discrimination by use of spatial frequency parameters in ultrasound B-mode images

The structural characteristics of a healthy tendon are related to the anisotropic speckle patterns observed in ultrasonic images. This speckle orientation is disrupted upon damage to the tendon structure as observed in patients with tendinopathy. Quantification of the structural appearance of tendon shows promise in creating a tool for diagnosing, prognosing, or measuring changes in tendon organization over time. The current work describes a first step taken towards this goal-classification of Achilles tendon images into tendinopathy and control categories. Eight spatial frequency parameters were extracted from regions of interest on tendon images, filtered and classified using linear discriminant analysis. Resulting algorithms had better than 80% accuracy in categorizing tendon image kernels as tendinopathy or control. Tendon images categorized wrongly provided for an interesting clinical association between incorrect classification of tendinopathy kernels as control and the symptom and clinical time history based inclusion criteria. To assess intersession reliability of image acquisition, the first 10 subjects were imaged twice during separate sessions. Test-retest of repeated measures was excellent (tau = 0.996, ICC = (2, 1) = 0.73 with one outlier) indicating a general consistency in imaging techniques.