Kevin D. Donohue

Genetic analysis in the Collaborative Cross breeding population

Genetic reference populations in model organisms are critical resources for systems genetic analysis of disease related phenotypes. The breeding history of these inbred panels may influence detectable allelic and phenotypic diversity. The existing panel of common inbred strains reflects historical selection biases, and existing recombinant inbred panels have low allelic diversity. All such populations may be subject to consequences of inbreeding depression. The Collaborative Cross (CC) is a mouse reference population with high allelic diversity that is being constructed using a randomized breeding design that systematically outcrosses eight founder strains, followed by inbreeding to obtain new recombinant inbred strains. Five of the eight founders are common laboratory strains, and three are wild-derived. Since its inception, the partially inbred CC has been characterized for physiological, morphological, and behavioral traits. The construction of this population provided a unique opportunity to observe phenotypic variation as new allelic combinations arose through intercrossing and inbreeding to create new stable genetic combinations. Processes including inbreeding depression and its impact on allelic and phenotypic diversity were assessed. Phenotypic variation in the CC breeding population exceeds that of existing mouse genetic reference populations due to both high founder genetic diversity and novel epistatic combinations. However, some focal evidence of allele purging was detected including a suggestive QTL for litter size in a location of changing allele frequency. Despite these inescapable pressures, high diversity and precision for genetic mapping remain. These results demonstrate the potential of the CC population once completed and highlight implications for development of related populations.

Estimating mean scatterer spacing with the frequency-smoothed spectral autocorrelation function

The quasiperiodicity of regularly spaced scatterers results in characteristic patterns in the spectra of backscattered ultrasonic signals from which the mean scatterer spacing can be estimated. The mean spacing has been considered for classifying certain biological tissue. This paper addresses the problem of estimating the mean scatterer spacing from backscattered ultrasound signals using the frequency-smoothed spectral autocorrelation (SAC) function. The SAC function exploits characteristic differences between the phase spectrum of the resolvable quasiperiodic scatterers and the unresolvable uniformly distributed (diffuse) scatterers to improve estimator performance over other estimators that operate directly on the magnitude spectrum. Mean scatterer spacing estimates are compared for the frequency-smoothed SAC function and a cepstral technique using an AR model. Simulation results indicate that SAC-based estimates converge more reliably over smaller amounts of data than cepstrum-based estimates. An example of computing an estimate from liver tissue scans is also presented for the SAC function and the AR cepstrum.<<ETX>>

Mean-scatterer spacing estimates with spectral correlation.

An ultrasonic backscattered signal from material comprised of quasiperiodic scatterers exhibit redundancy over both its phase and magnitude spectra. This paper addresses the problem of estimating mean-scatterer spacing from the backscattered ultrasound signal using spectral redundancy characterized by the spectral autocorrelation (SAC) function. Mean-scatterer spacing estimates are compared for techniques that use the cepstrum and the SAC function. A -scan models consist of a collection of regular scatterers with Gamma distributed spacings embedded in diffuse scatterers with uniform distributed spacings. The model accounts for attenuation by convolving the frequency dependent scattering centers with a time-varying system response. Simulation results indicate that SAC-based estimates converge more reliably over smaller amounts of data than cepstrum-based estimates. A major reason for the performance advantage is the use of phase information by the SAC function, while the cepstrum uses a phaseless power spectral density that is directly affected by the system response and the presence of diffuse scattering (speckle). An example of estimating the mean-scatterer spacing in liver tissue also is presented.

TISSUE CLASSIFICATION WITH GENERALIZED SPECTRUM PARAMETERS

This paper presents performance comparisons between breast tumor classifiers based on parameters from a conventional texture analysis (CTA) and the generalized spectrum (GS). The computations of GS-based parameters from radiofrequency (RF) ultrasonic scans and their relationship to underlying scatterer properties are described. Clinical experiments demonstrate classifier performances using 22 benign and 24 malignant breast mass regions taken from 40 patients. Linear classifiers based on parameters from the front edge, back edge and interior tumor regions are examined. Results show significantly better performances for GS-based classifiers, with improvements in empirical receiver operating characteristic (ROC) areas of greater than 10%. The ROC curves show GS-based classifiers achieving a 90% sensitivity level at 50% specificity when applied to the back-edge tumor regions, an 80% sensitivity level at 65% specificity when applied to the front-edge tumor regions, and a 100% sensitivity level at 45% specificity when applied to the interior tumor regions.

ROC analysis of ultrasound tissue characterization classifiers for breast cancer diagnosis

Breast cancer diagnosis through ultrasound tissue characterization was studied using receiver operating characteristic (ROC) analysis of combinations of acoustic features, patient age, and radiological findings. A feature fusion method was devised that operates even if only partial diagnostic data are available. The ROC methodology uses ordinal dominance theory and bootstrap resampling to evaluate A/sub z/ and confidence intervals in simple as well as paired data analyses. The combined diagnostic feature had an A/sub z/ of 0.96 with a confidence interval of [0.93, 0.99] at a significance level of 0.05. The combined features show statistically significant improvement over prebiopsy radiological findings. These results indicate that ultrasound tissue characterization, in combination with patient record and clinical findings, may greatly reduce the need to perform biopsies of benign breast lesions.

Backpropagation neural network design and evaluation for classifying weed species using color image texture

Color co-occurrence method (CCM) texture statistics were used as input variables for a backpropagation (BP) neural network weed classification model. Thirty-three unique CCM texture statistic inputs were generated for 40 images per class, within a six class data set. The following six classes were studied: giant foxtail, large crabgrass, common lambsquarter, velvetleaf, ivyleaf morningglory, and clear soil surface. The texture data was used to build six different input variable models for the BP network, consisting of various combinations of hue, saturation, and intensity (HSI) color texture statistics. The study evaluated classification accuracy as a function of network topology, and training parameter selection. In addition, training cycle requirements and training repeatability were studied. The BP topology evaluation consisted of a series of tests on symmetrical two hidden-layer network, a test of constant complexity topologies, and tapered topology networks. The best symmetrical BP network achieved a 94.7% classification accuracy for a model consisting of 11 inputs, five nodes at each of the two hidden layers and six output nodes (11 ×5 ×5 ×6 BP network). A tapered topology ( 11 ×12 ×6 ×6 BP network) out performed all other BP topologies with an overall accuracy of 96.7% and individual class accuracies of 90.0% or higher.

Analysis of order statistic filters applied to ultrasonic flaw detection using split-spectrum processing

Split-spectrum processing of broadband ultrasonic signals coupled with order statistic filtering has proven to be effective in improving the flaw-to-clutter ratio of backscattered signals. It is shown that an optimal rank can be obtained with a prior knowledge of flaw-to-clutter ratio and the underlying distributions. The order statistic filter performs well where the flaw and clutter echoes have good statistical separation in a given quantile region representing a particular rank (e.g. minimum, median, maximum). Order statistic filters are analyzed for the situation in which the observations do not contain equivalent statistical information. Experimental and simulated results are presented to show how effectively the order statistic filter can utilize information contained in different frequency bands to improve flaw detection.<<ETX>>

Assessment of a non-invasive high-throughput classifier for behaviours associated with sleep and wake in mice

This work presents a non-invasive high-throughput system for automatically detecting characteristic behaviours in mice over extended periods of time, useful for phenotyping experiments. The system classifies time intervals on the order of 2 to 4 seconds as corresponding to motions consistent with either active wake or inactivity associated with sleep. A single Polyvinylidine Difluoride (PVDF) sensor on the cage floor generates signals from motion resulting in pressure. This paper develops a linear classifier based on robust features extracted from normalized power spectra and autocorrelation functions, as well as novel features from the collapsed average (autocorrelation of complex spectrum), which characterize transient and periodic properties of the signal envelope. Performance is analyzed through an experiment comparing results from direct human observation and classification of the different behaviours with an automatic classifier used in conjunction with this system. Experimental results from over 28.5 hours of data from 4 mice indicate a 94% classification rate relative to the human observations. Examples of sequential classifications (2 second increments) over transition regions between sleep and wake behaviour are also presented to demonstrate robust performance to signal variation and explain performance limitations.

Analysis and classification of tissue with scatterer structure templates

Back-scattered ultrasonic signals provide scatterer structure information. Large-scale structures, such as tissue and tumor boundaries, typically create significant amplitude differences that reveal boundaries in conventional intensity images. Small-scale structures typically result in textures observed over regions of the intensity image. This paper describes the generalized spectrum (GS) for characterizing small-scale scatterer structures and applies it to analyze scatterer structures in a class of malignant and benign breast masses. Methods are presented for scaling and normalizing the GS to reduce effects from system response, overlaying tissue, and variability from noncritical structures. Results from a limited clinical study demonstrate an application of using the GS to discriminate between benign and malignant breast masses that contain internal echoes. Sections of rf A-scans in 41 breast mass regions were taken from 26 patients. A GS analysis was applied to determine critical structural properties between a class of fibroadenoma and carcinoma masses. Classifiers designed using significant structure differences identified by the GS analysis achieved approximately 82% true-positive and 10% false-positive rates.

Spectral correlation in ultrasonic pulse echo signal processing

The effects of using spectral correlation in a maximum-likelihood estimator (MLE) for backscattered energy corresponding to coherent reflectors embedded in media of microstructure scatterers is considered. The spectral autocorrelation (SAC) function is analyzed for various scatterer configurations based on the regularity of the interspacing distance between scatterers. It is shown that increased regularity gives rise to significant spectral correlation, whereas uniform distribution of scatters throughout a resolution cell results in no significant correlation between spectral components. This implies that when a true uniform distribution for the effective scatterers exists, the power spectral density (PSD) is sufficient to characterize their echoes. However, as the microstructure scatterer distribution becomes more regular, SAC terms become more significant. MLE results for 15 A-scans from stainless steel specimens with three different grain sizes indicate an average 6-dB signal-to-noise ratio (SNR) improvement in the coherent scatterer (flat-bottom hole) echo intensities for estimators using the SAC characterization as opposed to the PSD characterization.<<ETX>>