B-I-29
Towards Automated Digital Histopathology with Circularly Polarized Light
A. Bykov1, M. Borovkova1, V. Dremin2, O. Sieryi1, I. Meglinski1'2'3
1- Optoelectronics and Measurement Techniques Unit, University of Oulu, Oulu, Finland
2- College of Engineering and Physical Sciences, Aston University, Birmingham, UK
3- Institute of Clinical Medicine, I.M. Sechenov First Moscow State Medical University, Moscow, Russia
alexander.bykov@oulu.fi
The gold standard of tissue analysis is currently represented by the microscopic, slide-based procedure. Conventionally, the tissue preparation protocol requires the specimen to be formalin-fixed, dehydrated and embedded in paraffin wax to enable slicing into thin layers. After that, the obtained slices are chemically stained to contrast tissue constituents for microscopic inspection. The described procedure, together with the subsequent analysis, can take up to several weeks. Moreover, the possible sampling limitations and the inability to observe tissue organelles in the native environment limit the efficiency of the method. Thus, there is a need for slide-free, stain-free, non-destructive automated digital histopathological imaging to address the mentioned issues.
Due to the high sensitivity of polarized light to the variations of structure in the analyzed sample, the spatially resolved polarized light spectroscopy has a great potential to be used for screening malformations in biological tissues. Our studies show that circularly polarized light scattered within the tissues is highly sensitive to the presence of cancer cells [1] as well as Ap plaques [2] that allows identification of the cancerous lesions on early-stage or reliable Alzheimer's disease (AD) detection. For the measurements, the laboratory-built double-axis system containing both illumination and detection channels has been used. The illumination channel contains the supercontinuum fiber laser (Leukos Ltd., France). A high-speed acousto-optic tunable filter (Leukos Ltd., France) enables selecting the specific probing wavelength within the range of 400-650 nm. Circularly polarized light of the selected wavelength is then focused onto the sample at 55° angle. The light diffusely reflected from the sample is collected in the detection channel at 30° angle at a variable distance away from the point of incidence and analyzed by the Stokes polarimeter system (Thorlabs Ltd., USA). Spatial scanning of the samples is performed with the XY translation stage, providing a resolution of 5 ^m.
The multiple measurements have been performed on a formalin-fixed, paraffin-embedded blocks of early-stage breast cancer samples (ductal carcinoma in situ). We show that circularly polarized light scattered within the breast sample is sensitive to the presence of cancer cells and provides the information close to that obtained from the standard histological examination. The highest contrast between cancerous and healthy regions was observed for the wavelength of 450 nm. The degree of polarization (DoP) of the reflected light was found to be the most sensitive parameter indicating the malignant changes in the considered samples. Cluster analysis based on the k-means algorithm has been performed to implement the automated classification of the obtained data and delineate cancer zones of the specimens. A good correlation of the automatically recognized zones with the histopathological image obtained by a qualified pathologist with the conventional staining/slicing technique is observed.
The pilot results demonstrating the possibility of AD detection has been obtained in the mouse model. The comparison of the measured polarization parameters for the mice in the early stage of AD (group I, 11 animals, 50-75 days old) and for the well-developed disease (group II, 13 animals, 175-200 days old) is performed. The presence and the density of Ap plaques were also confirmed with the conventional histopathologic analysis. The DoP for the samples of group I was remarkably higher than that of group II that can be explained by the higher scattering of the tissue from group II. The higher magnitude of birefringence was observed for group II and can be explained by the growing presence of Ap plaques that exhibit significant birefringence. The proposed methodology demonstrates a great potential for developing a new diagnostic protocol for routine clinical practice.
The research was supported by the Academy of Finland (grants: 314369 and 325097) and EU H2020 ATTRACT project (777222). Authors acknowledge Northern Finland Biobank Borealis and Prof. Jens Pahnke, University of Oslo, Norway for the provided tissue samples.
[1] D. Ivanov et al., J. Biophoton. 13(8): e202000082 (2020).
[2]. M. Borovkova et al., Biomed. Opt. Exp. 11(8), 4509-4519 (2020).