Investigation of fluorescence lifetime imaging for quantitative cervical cancer screening and diagnosis

• Main Author: Gu, Jun
• Other Authors: Ng Beng Koon
• Format: Theses and Dissertations
• Language: English
• Published: 2015
• Subjects: DRNTU::Engineering::Electrical and electronic engineering::Antennas, wave guides, microwaves, radar, radio
• Online Access: https://hdl.handle.net/10356/64895
• Institution: Nanyang Technological University

To supplement the traditional histopathological examination of early cervical cancer, fluorescence lifetime imaging microscopy (FLIM) of H&E stained cervical tissue sections was investigated. Categories comprising normal, cervical intraepithelial neoplasia (CIN1, CIN2, CIN3), squamous cell carcinoma (SCC) and microinvasive were imaged and analyzed. Lifetime calculated from fluorescence emission decay in cervical epithelium region was used to correlate with tissue pathology. Mean and standard deviation of the characteristic fluorescence lifetime were used as feature vectors to feed into a neural network extreme learning machine (ELM) classifier for discrimination. An averaged sensitivity and specificity of 92.8% and 80.2%, respectively, were achieved when differentiating normal from precancerous (CIN1, CIN2, CIN3) samples. Discriminations between low-risk group (normal, CIN1) and high-risk group (CIN2, CIN3) were also conducted due to the clinical importance for treatment and a sensitivity of 85.0% and specificity of 87.6% were obtained. The results suggest that the proposed fluorescence lifetime imaging of H&E stained cervical tissue sections with the quantitative neural network ELM could be used to aid and supplement the traditional histopathological examinations. Epithelial layer analysis of cervical fluorescence lifetime images was investigated to improve the diagnosis of H&E stained cervical tissue sections. The previously defined ROIs in cervical epithelium regions were divided into even layers in the tissue growth direction and diagnostic value in each of the divided layers was explored. It was found that feature vectors including a combination of the characteristic lifetime from the top-half epithelium gave the maximum diagnostic accuracy. Neural network ELM classifier was used for discrimination and an averaged sensitivity of 94.6% and specificity of 84.3% were obtained when differentiating normal from precancerous cervical tissues. The proposed layer analysis technique has the advantage of achieving a concurrently higher sensitivity and specificity. In addition, analyzing only the top-half of the cervical epithelium shortens the calculation time. The proposed method can provide more accurate and faster cervical diagnosis, which can better supplement traditional gold standard histopathological examination. Liquid based cytological (LBC) Papanicolaou test is a very effective pre-invasive tool to screen for cervical intraepithelial neoplasia (CIN). However, LBC has limited sensitivity and specificity. Fluorescence lifetime imaging modality was applied to study H&E stained LBC cells. The resulting lifetime images were processed to extract the whole cell and cytoplasmic regions for analysis. The result shows that whole cell fluorescence lifetime distribution is highly similar to the cytoplasmic region fluorescence lifetime distribution. The mean of the characteristic fluorescence lifetime in the whole cell region was used to differentiate between normal and precancerous cells. A concurrently high sensitivity and specificity of 92.6% and 71.2% were achieved, respectively. This preliminary study suggests that the proposed method can potentially supplement and improve the automated screening of cervical LBC samples. The origin to the correlation between fluorescence lifetime changes and the associated cervical pathological states was investigated. Five proteins including BSA, poly-L-arginine, poly-L-histidine, poly-L-lysine and poly-L-tryptophan were used to simulate the proteins in cervical cytoplasm. Eosin Y, which is the only fluorophore involved in H&E staining, were used to bind with these proteins. Fluorescence lifetimes from the bound protein-dye complex were measured and compared. It was found that fluorescence lifetime varied with the solution pH value and environment solvent. The results obtained are in good agreement with the lifetime shortening trend observed in H&E stained cervical tissue sections and the lifetime lengthening behavior found in LBC cells.