Please use this identifier to cite or link to this item:
|Title:||Development of fast spectroscopic imaging techniques for biomedical applications||Authors:||Chen, Shuo||Keywords:||DRNTU::Engineering::Electrical and electronic engineering::Optics, optoelectronics, photonics||Issue Date:||2015||Source:||Chen, S. (2015). Development of fast spectroscopic imaging techniques for biomedical applications. Doctoral thesis, Nanyang Technological University, Singapore.||Abstract:||This dissertation presents a series of studies in the development of fast spectroscopic imaging techniques, including diffuse reflectance, fluorescence and Raman spectroscopic imaging, in which the biomedical applications of the techniques are demonstrated. First, the background of various optical spectroscopy techniques and the current state of art in spectroscopic imaging are presented, in which the new approach of narrow-band measurements followed by spectral reconstruction is introduced. As a critical step in the new approach, spectral reconstruction based on Wiener estimation have been investigated, which yielded several methods. In particular, the modified Wiener estimation (WE) method was developed to improve the reconstruction accuracy of the diffuse reflectance from narrow-band color measurements. It was demonstrated that the proposed modified WE can reconstruct diffuse reflectance spectra with higher accuracy than the traditional WE method. These diffuse reflectance spectra could be then used to estimate optical properties and further tissue parameters. Then a sequential weighted WE method was developed to derive tissue parameters directly from narrow-band color measurements and their ratios without reconstructing diffuse reflectance spectra. It was found that the sequential weighted WE method showed significant improvement in the accuracy of derived tissue parameters compared with the traditional WE method in a phantom experiment. The direct extraction of tissue parameters could facilitate the monitoring of tissue parameters in a large region of interest in real time for clinical diagnosis. Inspired by the above phantom experiment, narrow-band measurements and their ratios were used for the early prediction of flap occlusion in an animal study. Besides diffuse reflectance imaging, narrow-band autofluorescence imaging was investigated as well. The results showed the high feasibility of using narrow-band imaging to monitor flap occlusion. We further extend this spectral imaging technique to Raman spectroscopic imaging. The major challenge in Raman imaging is that the Raman signal in biological samples is intrinsically weak, and a Raman spectrum is usually more complex in spectral features such as the number of peaks than diffuse reflectance and fluorescence spectra. Another challenge is that there is no commercial polychromatic camera designed for Raman imaging. We addressed these challenges by applying narrow-band measurements to improve the signal-to-noise ratio and selecting/designing filters for a virtual Raman camera. The results of spectral reconstruction for Raman spectra both with and without fluorescence background showed excellent agreement with measured spectra, which implies the feasibility of using our approach of fast Raman imaging to investigate dynamically changing phenomena in biological samples. This technique can be also used to recover Raman spectra from low signal-to-noise ratio (SNR) Raman measurements. In this application, a low SNR Raman spectrum is integrated along the wavenumber dimension to reduce the influence of noise, which is followed by spectral reconstruction based on WE to recover the Raman spectrum with high spectral resolution. This approach showed the ability of recovering Raman spectra from measurements with extremely low SNR, which was more accurate than four commonly used de-noising methods. Despite the many advantages and applications of narrow-band measurements followed by spectral reconstruction, one major limitation of this technique is that a new calibration data is required for each type of samples, which implies a huge burden and may prevent this Raman imaging approach from being widely adopted. To overcome this limitation, we proposed a method to create a universal calibration dataset for spectral reconstruction. Because many biological samples, such as human cells, share the same set of basic biochemical components, the calibration dataset based on these biochemical components will be applicable to all such samples. In this case, only a handful number of Raman measurements are needed to create such a universal calibration dataset. Moreover, the measurements of those basic components can be reused if they are shared by a new category of samples. Therefore, the resources required for creation of the calibration dataset can be dramatically reduced using this new method. In summary, the proposed spectroscopic imaging technique, which refers to narrow-band measurements followed by spectral reconstruction, is able to realize fast spectral imaging or the quick extraction of key tissue parameters in the cases of diffuse reflectance, fluorescence and Raman spectroscopy or improve the signal-to-noise ratio of optical spectra from low-SNR measurements. The principle component (PC) based filters applied in the step of narrow-band measurements are found to generate the best performance. A new method has been developed to reduce the resources required for the creation of the calibration dataset in the step of spectral reconstruction. These techniques will be further refined and explored to observe fast changing phenomenon in biomedical applications.||URI:||https://hdl.handle.net/10356/65408||DOI:||10.32657/10356/65408||Fulltext Permission:||open||Fulltext Availability:||With Fulltext|
|Appears in Collections:||SCBE Theses|
Updated on May 11, 2021
Updated on May 11, 2021
Items in DR-NTU are protected by copyright, with all rights reserved, unless otherwise indicated.