ABSTRACT
In this paper, we investigate novel low-dimensional and model-free representations for multi-spectral fluorescence lifetime imaging microscopy (m-FLIM) data. We depart from the classical definition of the phasor in the complex plane to propose the extended output phasor (EOP) and extended phasor (EP) for multi-spectral information. The frequency domain properties of the EOP and EP are analytically studied based on a multiexponential model for the impulse response of the imaged tissue. For practical implementations, the EOP is more appealing since there is no need to perform deconvolution of the instrument response from the measured m-FLIM data, as in the case of EP. Our synthetic and experimental evaluations with m-FLIM datasets of human coronary atherosclerotic plaques show that low frequency indexes have to be employed for a distinctive representation of the EOP and EP, and to reduce noise distortion. The tissue classification of the m-FLIM datasets by EOP and EP also improves with low frequency indexes, and does not present significant differences by using either phasor.
ABSTRACT
Melanoma is the most aggressive skin cancer type. It is characterized by pigmented lesions with high tissue invasion and metastatic potential. The early detection of melanoma is extremely important to improve patient prognosis and survival rate, since it can progress to the deadly metastatic stage. Presently, the melanoma diagnosis is based on the clinical analysis of the macroscopic lesion characteristics such as shape, color, borders following the ABCD rules. The aim of this study is to evaluate the time-resolved fluorescence lifetime of NADH and FAD molecules to detect cutaneous melanoma in an experimental in vivo model. Forty-two lesions were analyzed and the data was classified using linear discriminant analysis, a sensitivity of 99.4%, specificity of 97.4% and accuracy of 98.4% were achieved. These results show the potential of this fluorescence spectroscopy for melanoma detection.