Sunyoung Park

Plasmon resonance coupling of metal nanoparticles for molecular imaging of carcinogenesis in vivo

An effective cancer control strategy requires improved early detection methods, patient-specific drug selection, and the ability to assess response to targeted therapeutics. Recently, plasmon resonance coupling between closely spaced metal nanoparticles has been used to develop ultrasensitive bioanalytical assays in vitro. We demonstrate the first in vivo application of plasmon coupling for molecular imaging of carcinogenesis. We describe molecular-specific gold bioconjugates to image epidermal growth factor receptor (EGFR); these conjugates can be delivered topically and imaged noninvasively in real time. We show that labeling with gold bioconjugates gives information on the overexpression and nanoscale spatial relationship of EGF receptors in cell membranes, both of which are altered in neoplasia. EGFR-mediated aggregation of gold nanoparticles in neoplastic cells results in more than a 100-nm color shift and a contrast ratio of more than tenfold in images of normal and precancerous epithelium in vivo, dramatically increasing contrast beyond values reported previously for antibody-targeted fluorescent dyes.

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.

Estimation of the Multivariate Multiple-Membership Random Effects Model With and Without Missing Data

In this study, we introduce the multivariate multiplemembership random effects model (MV-MMREM) for handling dependence resulting from analyzing multiple, related outcomes per individual when individuals are members of multiple clusters. Addressing multiplemembership helps avoid negatively impacting the estimation of random effects variance components. The use of multivariate models also improves estimation and power with missing data. Using real data, we compared (a) MV-MMREM estimates to those resulting from (b) the use of a multivariate hierarchical linear model (MV-HLM) that ignored the multiple-membership data structure and (c) the separate estimation of three univariate-MMREMs that ignored the dependency of the three outcomes. We initially compared the three models’ estimates using a subset of the ECLS-K dataset without missing data. We then introduced missing at random (MAR) data into two of three outcomes to provide a MAR dataset. We included all students with school identifiers for 1st, 3rd and 5th grades with complete reading, math, and science scores in 5th grade (resulting in data for 10,366 students from 1,856 schools). 1,192 students attended more than two schools during these three grades. For the MARdataset, 20% of math and 20% of science scores were deleted (lower reading scores led to higher probability of missingness of math and science scores). All models