Georgia Georgiou

Tissue characterization using the continuous wavelet transform. II. Application on breast RF data

For pt.I see ibid., vol.48, no.2, p.356-63 (2001). In the first part of this work (Georgiou and Cohen 2000), a wavelet-based decomposition algorithm of the RF echo into its coherent and diffuse components was introduced. In this paper, the proposed algorithm is used to estimate structural parameters of the breast tissue such as the number and energy of coherent scatterers, the energy of the diffuse scatterers, and the correlation between them. Based on these individual parameters, breast tissue characterization is performed. The database used consists of 155 breast scans from 42 patients. The results are presented in terms of empirical receiver operating characteristics (ROC) curves. The results of this study are discussed in relation to the tissue microstructure. Individual estimated parameters are able to differentiate reliably between normal and fibroadenoma or fibrocystic or cancerous tissue (area under the ROC A/sub z/>0.93). Also, the differentiation between malignant and benign (normal, fibrocystic, and fibroadenoma) tissue was possible (A/sub z/>0.89).

Is early detection of liver and breast cancers from ultrasound scans possible

This paper presents an integral approach for the tissue characterization problem. Such an approach includes a model, estimation algorithms and an evaluation method. This work focuses on liver and breast tissue characterization but it may be applicable to other tissue types after proper modifications. Liver and breast tissue is composed of two major kinds of scattering structure, i.e., the liver and breast parenchyma, which is relatively large and thus resolvable using the current ultrasonic transducers, and liver and breast cells which are not resolvable. In this work, we propose a decomposition approach for the RF echo into two components, namely the coherent and diffuse component, which are related to the resolvable and unresolvable scatterers in the liver and breast structure, respectively. Structural differences between the liver and breast, related to the resolvable scatterers properties, led us to develop two different decomposition algorithms. The first algorithm was developed for the liver RF echo and was based on the quasi-periodic structure of the liver lobules. Breast tissue decomposition was based on a more general model for the resolvable scatterers echo, because the breast tissue parenchyma is far from regular. By using the proposed decomposition we were able to estimate structural parameters of the liver and breast such as the average spacing of the liver lobules, the energy of the resolvable and unresolvable scatterers, and the correlation between neighboring unresolvable scatterers in the tissue. Empirical receiver operating characteristics analysis was applied to the parameters estimated from a large database of liver and breast B-scan images, to evaluate their diagnostic power. Single parameters of the liver and breast tissue showed good discriminating power between cancerous and normal liver and breast tissue, and also between malignant and benign breast tissue. The ability to identify small breast lesions (4 mm) is also demonstrated.

Unsupervised segmentation of RF echo into regions with different scattering characteristics

Recent experimental results verify that the probability distribution function of the diffuse component of the RF echo depends primarily on the concentration of the diffuse scatterers in the resolution cell. In this paper we apply these results to develop an unsupervised segmentation scheme that partitions an RF A-scan or B-scan image into statistically homogeneous regions that reflect the underlying scattering characteristics. The proposed segmentation scheme is based on a nonparametric homogeneity test that compares two regions of interest (ROI) for possible merging utilizing information about both the coherent and the diffuse component of the RF echo. For the coherent component, homogeneity is defined in terms of the estimated average spacing of each ROI. For the diffuse component, we use the nonparametric Kolmogorov-Smirnov (K-S) homogeneity statistical test that compares two empirical distributions associated with any two ROIs. This test can be used to obtain a segmentation into regions with different scattering characteristics regardless of the nature of the scattering conditions (e.g., Rayleigh regions with different scatterer concentration, different non-Rayleigh regions, or different coherent scattering regions). Finer segmentation can be obtained by learning the distributions associated with the various homogeneous regions obtained from the coarse segmenter. The proposed segmentation scheme is applied on simulated RF scans with different scatterer concentration per resolution cell, on phantom data which mimic tissue, and on liver scans. The results demonstrate the effectiveness of the segmentation algorithm even in cases of subtle differences in the scattering characteristics of each region (for example, diffuse component with scatterer density of 16 and 32 scatterers per resolution cell).

Detecting and estimating structure regularity of soft tissue organs from ultrasound images

This paper deals with the problem of detecting and estimating the scatterer spacing between the regularly-spaced resolvable coherent scatterers found in tissue. This parameter has been successfully used in classifying tissue structure, in differentiating between normal and cirrhotic liver, and in detecting diffuse liver disease. This paper presents a Wold decomposition of the radio frequency (RF) field into its diffused and coherent components from which maximum likelihood estimates (MLE) or minimum mean square error (MMSE) estimates of the scattering spacing is easily computed.

Malignant and benign breast tissue classification performance using a scatterer structure preclassifier

Benign and malignant breast tissue classification is examined for generalized-spectrum parameters computed from RF ultrasound data when a preclassification of subregions based on general scattering properties is performed. Results using a clinical database of 84 patients show statistically significant improvements (over 10% in receiver operation characteristic (ROC) areas) when only coherent scatterer subregions are used as compared to using all subregions within the region of interest.

Classification of simulated hyperplastic stages in the breast ducts based on ultrasound RF echo

Visual inspection of ultrasound is diagnostically limited for characterizing breast tissue, in particular when it comes to visually detecting hyperplasia that forms in the ducts at its early formation (at submillimeter resolution) stages. It can, of course, be seen using biopsies. But this will not be done unless the areas have been flagged using noninvasive modalities. The aim of this paper is to draw to the attention of the medical community (albeit through simulations) that the continuous wavelet transform decomposition (CWTD) that was proven in vivo for tissue characterization before has the potential to flag out simulated hyperplasia data at submillimeter resolutions. And it might be an excellent candidate for detecting in vivo hyperplastic changes in the breast. To the best of our knowledge, this is the first attempt at studying the potential of detecting cell growth in breast ducts using ultrasound. The stochastic decomposition model (the CWTD) of the RF echo with its coherent and diffuse components, yields image parameters that correlate closely with the structural parameters of the (simulated) hyperplastic stages of the breast tissue. The discrimination power of the various parameters is studied under a host of conditions, such as varying resolution, depth, and coherent to diffuse energy ratio (CDR) values using a point-scatterer model simulator that mimics epithelium hyperplastic growth in the breast ducts. These are shown to be useful for detecting the various types of simulated hyperplastic data. Careful analysis shows that three parameters, in particular the number of coherent scatterers, the Rayleigh scattering degree, and the energy of the diffuse scatterers, are most sensitive to variations in the hyperplastic simulated data. And they show very high ability to discriminate between various stages of simulated hyperplasia, even in cases of low resolution and low CDR values. Using the area under the receiver operating characteristics (ROC) curve (Az) as the performance metric, values of Az > 0.942 are obtained when discriminating between stages for resolution les 0.4 mm, even for low CDR values. Then it drops below the 0.9 range as the resolution exceeds the 0.4-mm range. A nonparametric segmentation method to extract ductal areas from breast scans is presented to be used as a pre-step before classification of hyperplastic stages in breast ducts. This is a necessary step for segmenting the RF scan into ductal versus nonductal areas from breast scans. This is tested using breast tissue mimicking phantom data resulting values of Az > 0.948 for different duct densities.

Detecting the stages of hyperplasia formation in the breast ducts using ultrasound B-scan images

A stochastic decomposition algorithm of the RF echo into its coherent and diffuse components is used towards estimating the structural parameters of the hyperplastic stages of the breast tissue leading to early breast cancer detection. The discrimination power of the various parameters is studied under a host of conditions such as varying resolution and SNR values using a point scatterer model simulator that mimics epithelium hyperplastic growth in the breast ducts. It is shown that three parameters, in particular, the number of coherent scatterers, the Rayleigh scattering degree and the energy of the diffuse scatterers, prove to show very high ability to discriminate between various stages of hyperplasia even in cases of low resolution and SNR values. Values of Az > 0.942 were obtained for resolution less than or equal to 0.4 mm even in low SNR values, then it drops below the 0.9 range as the resolution exceeds the 0.4 mm range

A Wold decomposition‐based autonomous system for detecting breast lesions in ultrasound images of the breast

This paper presents an autonomous system for detecting lesions in the breast. The Wold decomposition algorithm described is used to decompose the RF echo of the breast into its diffuse and coherent components. The coherent component is modeled as a periodic sequence and the diffuse component is modeled as an autoregressive time series of low order. The parameters of the model are estimated from selected regions of the RF image and used as detection features. The database of images that was used contained 370 B‐scan images from 52 patients, obtained in the Radiology department of the Thomas Jefferson Hospital. The pathologies of interest are carcinoma fibrocystic and stromal fibrosis disease and fibroadenoma. Empirical ROC techniques were used to evaluate the detection rate on single parameters of the model, such as the residual error variance and the autoregressive parameters of the diffuse component of the RF echo. The area under the empirical ROC curve for detecting lesion regions versus normal RF regio...

Can a machine outperform a clinician in interpreting ultrasound images

This paper deals with a method of detecting and estimating the scatterer spacing between the regularly spaced resolvable coherent scatterers in tissue. Scatterer spacing has been successfully used in classifying tissue structure, in differentiating between normal and cirrhotic liver, and in detecting diffuse liver disease. This paper presents a Wold decomposition of the radio frequency (rf) field into its diffused and coherent components from which maximum likelihood estimates (MLE) or minimum mean square error (MMSE) estimates of the scattering spacing are easily computed. The MLE are efficient and for relatively long record are unbiased. They result in accurate estimates in low signal-to-noise (SNR) ratios. Unfortunately, they require nonlinear minimization and knowledge of the probability density associated with the rf backscatter echo. The MMSE estimates, on the other hand, are computational simple, yield unique closed form solutions, do not require a priori knowledge of the probability distribution function of the backscatter echo, and result in accurate estimates in low signal-to-noise (SNR) ratios. The paper also presents an unbiased decision rule to detect whether or not an rf echo exhibits any specular scattering relative to the wavelength of the interrogating ultrasonic pulse. The approach has been tried on simulations as well as on in vivo scans of liver data, and appears to perform well.