Use of hyperspectral imaging to distinguish normal, precancerous, and cancerous cells

BACKGROUND. The objective of the current study was to test the hypothesis that the cytologic diagnosis of cancer cells can be enhanced by the technique of hyperspectral imaging (HSI). METHODS. As a proof of principle, HSI was employed to obtain hyperspectrum from a normal human fibroblast, as well as its telomerase-immortalized and SV40-transformed derivatives. Novel algorithms were developed to differentiate among these cell models based on spectral and spatial differences. Using the same technique with modified algorithms, the authors were able to differentiate among normal and precancerous (low-grade [LG] and high-grade [HG]) cervical cells and squamous cell carcinoma (SCC) on liquid-based Papanicolaou (Pap) test slides. RESULTS. The specificity for identifying normal fibroblast cell type based on spatial and spectral algorithms was 74.2%. The sensitivity for identifying telomerase-immortalized and SV40 -transformed cells was 100% and 90.3%, respectively. The system identified normal cervical cells with a specificity of 95.8%. With regard to LG precancerous cells and HG precancerous cells, the sensitivity was 66.7% and 93.5%, respectively. The sensitivity detected for SCC was 98.6%. CONCLUSIONS. HSI can be utilized in prescreening liquid-based Pap test slides to improve efficiency in Pap test diagnoses with the goal of ultimately reducing the mortality from cervical cancer while reducing health care costs.