TY - GEN
T1 - Automated Polarized Hyperspectral Imaging (PHSI) for ex-vivo and in-vivo Tissue Assessment
AU - Ma, Ling
AU - Srinivas, Akhila
AU - Krishnamurthy, Abirami
AU - Zhou, Ximing
AU - Shah, Nimit Subhashbhai
AU - Obaid, Girgis
AU - Fei, Baowei
N1 - Publisher Copyright:
© 2023 SPIE
PY - 2023
Y1 - 2023
N2 - Polarized light interactions with biological tissues can reveal information regarding tissue structure, while spectral characteristics are closely related to tissue composition. An integration of both modalities in a compact system could better assist tissue assessment. This study aims to develop a polarized hyperspectral imaging (PHSI) system that fulfills both linearly and circularly polarized hyperspectral imaging for in vivo and ex vivo applications. The system is comprised of a white LED, two linear polarizers, two liquid crystal variable retarders (LCVRs), and a hyperspectral snapshot camera. The system was calibrated to compute the full Stokes polarimetry. For tissue differentiation, fresh ex vivo mouse tissue specimens from kidney, liver, spleen, muscle, lung, and salivary gland of mice were imaged. The spectra of three features, named degree of polarization (DOP), degree of linear polarization (DOLP), and degree of circular polarization (DOCP), were generated. A k-nearest neighbor (k-NN) classifier was trained with multi-class spectra and 5-fold cross validation. It was found that DOP better differentiates tissue with an average accuracy of 0.87. Additionally, support vector machine (SVM) classifiers were trained to differentiate between each two of the organs, and it was determined that DOLP better identified kidney, liver, and spleen, whereas DOCP better identified muscle and lung tissues. Then, the setup was employed to image in vivo human fingers with and without a blood occlusion to qualitatively estimate oxygen saturation. Preliminary results demonstrate that both DOLP and DOCP reveal a distinction of oxygen saturation states. These results demonstrate the feasibility of the PHSI system for distinguishing between optical properties of tissues, which has the potential to reveal disease-related information for diverse medical applications.
AB - Polarized light interactions with biological tissues can reveal information regarding tissue structure, while spectral characteristics are closely related to tissue composition. An integration of both modalities in a compact system could better assist tissue assessment. This study aims to develop a polarized hyperspectral imaging (PHSI) system that fulfills both linearly and circularly polarized hyperspectral imaging for in vivo and ex vivo applications. The system is comprised of a white LED, two linear polarizers, two liquid crystal variable retarders (LCVRs), and a hyperspectral snapshot camera. The system was calibrated to compute the full Stokes polarimetry. For tissue differentiation, fresh ex vivo mouse tissue specimens from kidney, liver, spleen, muscle, lung, and salivary gland of mice were imaged. The spectra of three features, named degree of polarization (DOP), degree of linear polarization (DOLP), and degree of circular polarization (DOCP), were generated. A k-nearest neighbor (k-NN) classifier was trained with multi-class spectra and 5-fold cross validation. It was found that DOP better differentiates tissue with an average accuracy of 0.87. Additionally, support vector machine (SVM) classifiers were trained to differentiate between each two of the organs, and it was determined that DOLP better identified kidney, liver, and spleen, whereas DOCP better identified muscle and lung tissues. Then, the setup was employed to image in vivo human fingers with and without a blood occlusion to qualitatively estimate oxygen saturation. Preliminary results demonstrate that both DOLP and DOCP reveal a distinction of oxygen saturation states. These results demonstrate the feasibility of the PHSI system for distinguishing between optical properties of tissues, which has the potential to reveal disease-related information for diverse medical applications.
KW - Polarized hyperspectral imaging
KW - classification
KW - ex vivo
KW - in vivo
KW - oxygen saturation
KW - tissue assessment
UR - http://www.scopus.com/inward/record.url?scp=85159760799&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85159760799&partnerID=8YFLogxK
U2 - 10.1117/12.2651011
DO - 10.1117/12.2651011
M3 - Conference contribution
C2 - 38476292
AN - SCOPUS:85159760799
T3 - Progress in Biomedical Optics and Imaging - Proceedings of SPIE
BT - Label-free Biomedical Imaging and Sensing (LBIS) 2023
A2 - Shaked, Natan T.
A2 - Hayden, Oliver
PB - SPIE
T2 - Label-free Biomedical Imaging and Sensing (LBIS) 2023
Y2 - 28 January 2023 through 31 January 2023
ER -