Mia K. Markey

Complex Wavelet Structural Similarity: A New Image Similarity Index

We introduce a new measure of image similarity called the complex wavelet structural similarity (CW-SSIM) index and show its applicability as a general purpose image similarity index. The key idea behind CW-SSIM is that certain image distortions lead to consistent phase changes in the local wavelet coefficients, and that a consistent phase shift of the coefficients does not change the structural content of the image. By conducting four case studies, we have demonstrated the superiority of the CW-SSIM index against other indices (e.g., Dice, Hausdorff distance) commonly used for assessing the similarity of a given pair of images. In addition, we show that the CW-SSIM index has a number of advantages. It is robust to small rotations and translations. It provides useful comparisons even without a preprocessing image registration step, which is essential for other indices. Moreover, it is computationally less expensive.

Automated recognition of patterns characteristic of subcellular structures in fluorescence microscopy images

Methods for numerical description and subsequent classification of cellular protein localization patterns are described. Images representing the localization patterns of 4 proteins and DNA were obtained using fluorescence microscopy and divided into distinct training and test sets. The images were processed to remove out-of-focus and background fluorescence and 2 sets of numeric features were generated: Zernike moments and Haralick texture features. These feature sets were used as inputs to either a classification tree or a neural network. Classifier performance (the average percent of each type of image correctly classified) on previously unseen images ranged from 63% for a classification tree using Zernike moments to 88% for a backpropagation neural network using a combination of features from the 2 feature sets. These results demonstrate the feasibility of applying pattern recognition methods to subcellular localization patterns, enabling sets of previously unseen images from a single class to be classified with an expected accuracy greater than 99%. This will provide not only a new automated way to describe proteins, based on localization rather than sequence, but also has potential application in the automation of microscope functions and in the field of gene discovery.

Application of the mutual information criterion for feature selection in computer-aided diagnosis.

The purpose of this study was to investigate an information theoretic approach to feature selection for computer-aided diagnosis (CAD). The approach is based on the mutual information (MI) concept. MI measures the general dependence of random variables without making any assumptions about the nature of their underlying relationships. Consequently, MI can potentially offer some advantages over feature selection techniques that focus only on the linear relationships of variables. This study was based on a database of statistical texture features extracted from perfusion lung scans. The ultimate goal was to select the optimal subset of features for the computer-aided diagnosis of acute pulmonary embolism (PE). Initially, the study addressed issues regarding the approximation of MI in a limited dataset as it is often the case in CAD applications. The MI selected features were compared to those features selected using stepwise linear discriminant analysis and genetic algorithms for the same PE database. Linear and nonlinear decision models were implemented to merge the selected features into a final diagnosis. Results showed that the MI is an effective feature selection criterion for nonlinear CAD models overcoming some of the well-known limitations and computational complexities of other popular feature selection techniques in the field.

A machine learning perspective on the development of clinical decision support systems utilizing mass spectra of blood samples

Currently, the best way to reduce the mortality of cancer is to detect and treat it in the earliest stages. Technological advances in genomics and proteomics have opened a new realm of methods for early detection that show potential to overcome the drawbacks of current strategies. In particular, pattern analysis of mass spectra of blood samples has attracted attention as an approach to early detection of cancer. Mass spectrometry provides rapid and precise measurements of the sizes and relative abundances of the proteins present in a complex biological/chemical mixture. This article presents a review of the development of clinical decision support systems using mass spectrometry from a machine learning perspective. The literature is reviewed in an explicit machine learning framework, the components of which are preprocessing, feature extraction, feature selection, classifier training, and evaluation.

Differentiating tumor recurrence from treatment necrosis: a review of neuro-oncologic imaging strategies

Differentiating treatment-induced necrosis from tumor recurrence is a central challenge in neuro-oncology. These 2 very different outcomes after brain tumor treatment often appear similarly on routine follow-up imaging studies. They may even manifest with similar clinical symptoms, further confounding an already difficult process for physicians attempting to characterize a new contrast-enhancing lesion appearing on a patients follow-up imaging. Distinguishing treatment necrosis from tumor recurrence is crucial for diagnosis and treatment planning, and therefore, much effort has been put forth to develop noninvasive methods to differentiate between these disparate outcomes. In this article, we review the latest developments and key findings from research studies exploring the efficacy of structural and functional imaging modalities for differentiating treatment necrosis from tumor recurrence. We discuss the possibility of computational approaches to investigate the usefulness of fine-grained imaging characteristics that are difficult to observe through visual inspection of images. We also propose a flexible treatment-planning algorithm that incorporates advanced functional imaging techniques when indicated by the patients routine follow-up images and clinical condition.

Self-organizing map for cluster analysis of a breast cancer database

The purpose of this study was to identify and characterize clusters in a heterogeneous breast cancer computer-aided diagnosis database. Identification of subgroups within the database could help elucidate clinical trends and facilitate future model building. A self-organizing map (SOM) was used to identify clusters in a large (2258 cases), heterogeneous computer-aided diagnosis database based on mammographic findings (BI-RADS) and patient age. The resulting clusters were then characterized by their prototypes determined using a constraint satisfaction neural network (CSNN). The clusters showed logical separation of clinical subtypes such as architectural distortions, masses, and calcifications. Moreover, the broad categories of masses and calcifications were stratified into several clusters (seven for masses and three for calcifications). The percent of the cases that were malignant was notably different among the clusters (ranging from 6 to 83%). A feed-forward back-propagation artificial neural network (BP-ANN) was used to identify likely benign lesions that may be candidates for follow up rather than biopsy. The performance of the BP-ANN varied considerably across the clusters identified by the SOM. In particular, a cluster (#6) of mass cases (6% malignant) was identified that accounted for 79% of the recommendations for follow up that would have been made by the BP-ANN. A classification rule based on the profile of cluster #6 performed comparably to the BP-ANN, providing approximately 25% specificity at 98% sensitivity. This performance was demonstrated to generalize to a large (2177) set of cases held-out for model validation.

Automated image analysis of digital colposcopy for the detection of cervical neoplasia

Digital colposcopy is a promising technology for the detection of cervical intraepithelial neoplasia. Automated analysis of colposcopic images could provide an inexpensive alternative to existing screening tools. Our goal is to develop a diagnostic tool that can automatically identify neoplastic tissue from digital images. A multispectral digital colposcope (MDC) is used to acquire reflectance images of the cervix with white light before and after acetic-acid application in 29 patients. A diagnostic image analysis tool is developed to identify neoplasia in the digital images. The digital image analysis is performed in two steps. First, similar optical patterns are clustered together. Second, classification algorithms are used to determine the probability that these regions contain neoplastic tissue. The classification results of each patients images are assessed relative to the gold standard of histopathology. Acetic acid induces changes in the intensity of reflected light as well as the ratio of green to red reflected light. These changes are used to differentiate high-grade squamous intraepithelial (HGSIL) and cancerous lesions from normal or low-grade squamous intraepithelial (LGSIL) tissue. We report diagnostic performance with a sensitivity of 79% and a specificity of 88%. We show that diagnostically useful digital images of the cervix can be obtained using a simple and inexpensive device, and that automated image analysis algorithms show a potential to identify histologically neoplastic tissue areas.

Monte Carlo lookup table-based inverse model for extracting optical properties from tissue-simulating phantoms using diffuse reflectance spectroscopy

Abstract. We present a Monte Carlo lookup table (MCLUT)-based inverse model for extracting optical properties from tissue-simulating phantoms. This model is valid for close source-detector separation and highly absorbing tissues. The MCLUT is based entirely on Monte Carlo simulation, which was implemented using a graphics processing unit. We used tissue-simulating phantoms to determine the accuracy of the MCLUT inverse model. Our results show strong agreement between extracted and expected optical properties, with errors rate of 1.74% for extracted reduced scattering values, 0.74% for extracted absorption values, and 2.42% for extracted hemoglobin concentration values.

Assessment of Breast Aesthetics

Summary: A good aesthetic outcome is an important endpoint of breast cancer treatment. Subjective ratings, direct physical measurements, measurements on photographs, and assessment by three-dimensional imaging are reviewed and future directions in aesthetic outcome measurements are discussed. Qualitative, subjective scales have frequently been used to assess aesthetic outcomes following breast cancer treatment. However, none of these scales has achieved widespread use because they are typically vague and have low intraobserver and interobserver agreement. Anthropometry is not routinely performed because conducting the large studies needed to validate anthropometric measures (i.e., studies in which several observers measure the same subjects multiple times) is impractical. Quantitative measures based on digital/digitized photographs have yielded acceptable results but have some limitations. Three-dimensional imaging has the potential to enable consistent, objective assessment of breast appearance, including properties (e.g., volume) that are not available from two-dimensional images. However, further work is needed to define three-dimensional measures of aesthetic properties and how they should be interpreted.

3D Face Recognition Founded on the Structural Diversity of Human Faces

We present a systematic procedure for selecting facial fiducial points associated with diverse structural characteristics of a human face. We identify such characteristics from the existing literature on anthropometric facial proportions. We also present three dimensional (3D) face recognition algorithms, which employ Euclidean/geodesic distances between these anthropometric fiducial points as features along with linear discriminant analysis classifiers. Furthermore, we show that in our algorithms, when anthropometric distances are replaced by distances between arbitrary regularly spaced facial points, their performances decrease substantially. This demonstrates that incorporating domain specific knowledge about the structural diversity of human faces significantly improves the performance of 3D human face recognition algorithms.