ECE Theses and Dissertations

Permanent URI for this collection

https://hdl.handle.net/1911/8316

Browse

Browsing ECE Theses and Dissertations by Author "Appiah, Benjamin"

Now showing 1 - 2 of 2
  • Thumbnail Image
    Item
    Inverse model for the extraction of biological parameters from ovarian tissue fluorescence spectra and multivariate classifiers for tissue diagnosis
    (2007) Appiah, Benjamin; Drezek, Rebekah A.
    We present an inverse model to decompose bulk fluorescence spectra and extract tissue biological parameters. By deconvolving the effects of absorption and scattering from measured spectra, we are able to extract the intrinsic contributions from cellular and stromal fluorophores. Ovarian fluorescence, acquired ex-vivo immediately upon removal from patients, are analyzed using this model. We test the validity of the inverse model, and show that it has the ability to improve our understanding of the biological changes that cause the observed differences in the fluorescence spectra. The outputs of the model are used to develop classifiers for tissue diagnosis. Classifiers based on PLS selected features are also developed. An accuracy of more than 93% for discrimination between normal and cancerous ovarian tissues is achieved.
  • Thumbnail Image
    Item
    Optical spectroscopy and imaging systems for gynecological cancers: from Ultraviolet-C (UVC) to the Mid-infrared
    (2011) Appiah, Benjamin; Drezek, Rebekah A.
    Optical spectroscopy and imaging has proving to be of diagnostic relevance in many organ sites. We use fluorescence and FTIR spectroscopy to study gynecological organ sites and develop classification algorithms for cancer diagnosis. Ovarian cancer is the deadliest gynecological cancer. The American Cancer Society reports that for the year 20 I 0, there would be 21,880 new cases of ovarian cancer and 13,850 fatalities. This is partly due to the fact that current diagnostic and screening methods for the disease are not very accurate. In this study, we analyze the fluorescence spectra of excised normal and cancerous ovarian tissues at multiple excitation wavelengths. The data includes spectra obtained at the UVC wavelength 270nm and UVB wavelength 300nm. Excitation in the UVC has been especially understudied in spectroscopy for tissue diagnosis. We introduce the application of a novel SVM algorithm for the classification of fluorescence data. This SVM is trained subject to the Neyman Pearson (NP) criterion which allows for a decision rule that maximizes the detection specificity whilst constraining the sensitivity to a high value. This technique allows us to develop a binary classification algorithm that is not biased towards the larger group and this in tum leads to robust classifiers that are more suitable for clinical applications. We obtained sensitivities and specificities greater than 90% for ovarian cancer diagnosis using this algorithm. Also, FTIR is used to analyze cervical tissues. Absorption of light in the mid-IR region by biomolecules show up as peaks in the FTIR spectra, and there is differential absorption in tissue depending on the histopathology. The spectroscopic analysis informed our choosing of a wavelength for the illumination source ofa mid-IR microscope. We then present the design of an imaging system that employs the use ofa mid-IR quantum cascade laser(QCL) which can potentially have clinical use in the future. Finally a reflectance based fiber endoscope imaging system is presented. Cellular imaging is demonstrated with this system that has the potential for use in optical biopsy.