Hyperspectral Imaging to Discern Malignant and Benign Canine Mammary Tumors

Abstract

Hyperspectral imaging is an emerging technology in the field of biomedical engineering which may be used as a non-invasive modality to characterize tumors. In this thesis, a hyperspectral imaging system was used to characterize canine mammary tumors of unknown histopathology (pre-surgery) and correlate the results with the post-surgical histopathology results. The system consisted of a charge coupled device (CCD) camera, a liquid crystal tunable filter in the near infrared range (650-1100 nm), and a controller. Spectral signatures of malignant and benign canine mammary tumors were extracted and analyzed. The reflectance intensities of malignant tumor spectra were generally lower than benign tumor spectra over the wavelength range 650-1100nm. Previous studies have shown that cancerous tissues have a higher hemoglobin and water content, and lower lipid concentration with respect to benign tissues. The decreased reflectance intensity observed for malignant tumors is likely due to the increased microvasculature and, therefore, higher blood content of malignant tissue relative to benign tissue. Second derivative method was applied to the reflectance spectra. Peaks at 700, 840, 900 and 970 nm were observed in the second derivative reflectance spectra. These peaks were attributed to deoxy-hemoglobin, oxy-hemoglobin, lipid and water respectively. A Tissue Optical Index (TOI) was developed that enhances contrast between malignant and benign canine tumors. This index is based on the ratio of the reflectance intensity values corresponding to the wavelengths associated with the four chromophores. Principal Component Analysis (PCA) and Linear Discriminant Analysis (LDA) were also applied on the canine spectral dataset and the method was cross-validated. Preliminary results from 22 canine mammary tumors showed that the sensitivity and specificity of the PCA-LDA is method is 86% and 86% respectively. The sensitivity and specificity of the TOI model is 86% and 95% respectively. These results show promise in the non-invasive optical diagnosis of canine mammary cancer.