Multispectral Imaging Method for Rapid Identification and Analysis of Paraffin-Embedded Pathological Tissues

**Background:** Multi- and hyperspectral imaging are non-destructive techniques that integrate spectroscopy and imaging, used in medical applications such as disease diagnosis and image-guided surgery. These methods can detect changes in tissue scattering and absorption properties due to disease. Folic acid (vitamin B9) is essential for embryonic development, and maternal deficiency is linked to fetal abnormalities, including nervous system defects. Previous studies by the authors showed that maternal folic acid-deficient (FAD) diet alters protein expression and fiber/cell orientation in embryonic eye tissues. Traditional analysis requires staining or labeling, but this study aimed to develop a label-free multispectral imaging method to detect structural alterations in paraffin-embedded tissues as a diagnostic aid.

**Methods:** Eight-week-old C57/BL/6J female mice were assigned to three groups: control (standard diet, 2 mg folic acid/kg), FAD diet for 2 weeks (D2), and FAD diet for 8 weeks (D8). At 14.5 days of gestation (E14.5), embryos were removed, fixed, and embedded in paraffin. Frontal sections (5 µm thick) were prepared. Multispectral images were acquired using a tunable light source (390–705 nm, 15 nm steps) and a CCD camera, resulting in 22-band spectral data cubes. For each of the 27 eyes (9 per group), transmission images were obtained, and absorbance (optical density) was calculated. Principal component analysis (PCA) was applied to reduce dimensionality, and eigenvectors and eigenvalues were computed. Endmembers (pure spectral signatures) were derived from eigenvectors, and abundance maps were generated using non-negative least squares (NNLS) algorithm. Autofluorescence images were also captured using a FITC filter.

**Key Results:** PCA revealed that five principal components accounted for >99% of total variance in all groups. The first eigenvalue was 2.41 (control), 1.83 (D2), and 1.93 (D8), explaining 88.08%, 85.63%, and 86.79% of variance, respectively. Eigenvectors were similar across groups but showed slight spectral differences between control and FAD groups. The average absorbance peaked around 420 nm. Five endmembers were identified. Abundance maps for endmember #1 (associated with hemoglobin absorption, with peaks at 450–500 nm and 550 nm, and a drop after 600 nm) showed high abundance in vitreous and choroid (blood vessel regions) in controls, but decreased in D2 and D8 groups, indicating vascular alterations. Abundance maps for endmember #3 showed high abundance (>0.2) in retina and lens, suggesting microstructural changes. Autofluorescence images showed stronger green autofluorescence in D2 and D8 tissues compared to controls.

**Clinical Implications:** This multispectral imaging method can detect tissue alterations in paraffin-embedded, unstained samples, potentially serving as a rapid, non-invasive diagnostic tool for pathological tissues. The technique identified changes in blood vessel regions (vitreous and choroid) likely due to hemoglobin/oxyhemoglobin imbalance from folic acid deficiency, consistent with known effects on red blood cell metabolism and oxygen balance. It also detected microstructural changes in the retina and lens, possibly related to extracellular matrix organization. This method could reduce the need for staining or labeling in histopathological analysis, aiding in early diagnosis of developmental abnormalities or other ocular pathologies. However, further studies with more tissue types and optimized parameters are needed before clinical application.