Multi-spectral polarimetry imaging for early gastric cancer diagnosis

Abstract

Gastric cancer, which develops from the lining of stomach, is the fifth most severe disease worldwide. It leads to significant number of death each year. The early diagnosis of gastric cancer can dramatically increase the survival rate within five years. Thus the investigation on the early diagnosis of gastric cancer is of paramount importance. Polarimetry imaging, a technique utilizing the polarization properties of biological samples, is of great potential to diagnose cancer. This technique is able to provide quantitative information about tissue morphology and structural properties noninvasively. Thus, it is worth investigating the diagnosis of gastric cancer using polarimetry imaging. This dissertation presents several studies in the development of polarimetry imaging techniques for gastric cancer diagnosis in order to approach the eventual goal of the detection of early gastric cancer in vivo in real time. The background about gastric cancer, polarimetry imaging and polarimetry is first introduced. Then the comparison between cancer and normal gastric samples were performed at different wavelengths based on different polarization properties (retardance, diattenuation, depolarization, linear retardance, linear diattenuation, linear depolarization, circular retardance and circular depolarization) of samples together with all possible combinations. It was observed that all the polarization parameters except diattenuation and circular retardance demonstrated significant differences between cancer and normal samples for 4 m gastric samples. Moreover, it was found that wavelength did not play a critical role in improving the diagnostic accuracy. To further acquire more detailed information, we measured samples under 20X objective lens at only one wavelength 632 nm. Besides cancer and normal glands, other samples including intestinal metaplasia and dysplasia were also included. The data processing was conducted at the both macroscopic and microscopic level for different regions of interest ranging from 1×1 to 640×512. The results demonstrated the feasibility of performing “digital staining” using polarimetry imaging. While these results showed the great potential of polarimetry imaging in classifying gastric samples especially between cancer and normal, it is quite time consuming using the above systems and the equipment cannot be incorporated into a commercial endoscope. To address this problem, we proposed a snapshot polarimetry imaging method and the algorithm to enable all measurements in real time. The system has been tested on standard samples such as air and polarizers. The potential reasons for disagreement between experimental and theoretical results were discussed and future directions were pointed out. In summary, we have demonstrated that polarimetry imaging is of great potential to classify gastric samples at both the macroscopic and microscopic levels, especially for distinguishing between cancer and normal tissue samples. Snapshot polarimetry imaging is worth further investigation to achieve real time polarimetry imaging for clinical uses.