Utilising Hyperspectral Autofluorescence Imaging in the Objective Assessment of Disease State and Pain in Patients with Rheumatoid Arthritis

Abstract

Rheumatoid Arthritis (RA) is a chronic inflammatory disease resulting in joint swelling and pain. Treatment options can be reliant on disease activity scores (DAS) incorporating patient global assessments, which are quantified via visual analogue scales (VAS). VAS can be subjective and not necessarily align with clinical symptoms, such as inflammation, resulting in a disconnect between the patient's and practitioners' experience. The development of more objective assessments of pain would enable a more targeted and personalised management of pain within individuals with RA and have the potential to improve the reliability of assessments in research. Using emerging light-based hyperspectral autofluorescence imaging (HAI) technology, we aimed to objectively differentiate disease and pain states based on the analysis of synovial tissue (ST) samples from RA patients. In total, 22 individuals with RA were dichotomised using the DAS in 28-joint counts (DAS-28) into an inactive (IA) or active disease (active-RA) group and then three sub-levels of pain (low, mid, high) based on VAS. HAI was performed on ST sections to identify and quantify the most prominent fluorophores. HAI fluorophore analysis revealed a distinct separation between the IA-RA and active-RA mid-VAS cohort, successfully determining disease state. Additionally, the separation between active-RA Mid-VAS and active RA High-VAS cohort suggests that HAI could be used to objectively separate individuals based on pain severity.

Keywords: arthritis; hyperspectral; inflammation; rheumatoid.

Figure 1

(A), Steps of hyperspectral image acquisition and processing. A sample is illuminated by a given wavelength and emitted light resulting from the excitation is collected through filters coupled to a camera. A three-dimensional data stack (x,y,λ) comprises x-y images at different wavelengths (λ), which can then be interpreted for display. (B) Side-by-side comparison of an endocardial left atrial surface with three radiofrequency ablation under room light and a composite hyperspectral image of the same tissue, which clearly shows the borders of the lesion. Source: Adapted from Muselimyan et al. [18] under a cc attribution license.

Figure 2

Histological assessment of CD68 within ST. (A) Inactive RA disease state. (B) Active RA disease state. Representative samples used for images (A,B) were categorised based on the DAS-28 score. Representative images were obtained at 20× magnification. Semi-quantitative analysis of CD68 positive cells was performed and then categorised based on the (C) DAS-28 and (D) CRP values. Scale bars are representative of 100 µm. No statistically significant difference was observed in either grouping method. Error bars are SEM.

Figure 3

Unmixed auto fluorophore spectral signals from SC within ST samples obtained from inactive (IA) and active-RA patients. (A) Relative abundance of NAD(P)H, (B) relative abundance of Flavins, (C) relative abundance of collagen, and (D) the ratio of relative abundances of NAD(P)H to Flavin’s. Median of the relative abundances displayed as SEM. * p < 0.05, ** p < 0.01 and *** p < 0.001.

Figure 4

Differentiation of SC from IA and active-RA patients with varying disease and pain states via unmixing and autofluorescence feature analysis. (A) Clusters of IA and active cells, (B) clusters of IA-Low and IA-mid individuals based on SC, (C) clusters of RA-Mid and RA-High cells in the two groups, and (D) clustering of IA-Mid and RA-Mid cells. Symbols represent individual cells. Ellipses are one standard deviation around the mean, with the central point indicated by a cross.

Figure 5

Unmixed autofluorophore abundances for synovial fibrous tissue (FT). Samples were obtained from inactive (IA) and active-RA patients. (A) Relative abundance of NAD(P)H, (B) relative abundance of flavins, (C) relative abundance of collagen, and (D) the ratio of relative abundances of NAD(P)H and flavins. Median of the relative abundances displayed and SEM. * p < 0.05, and *** p < 0.001.

Figure 6

Differentiation of FT from inactive (IA) and active RA patients with varying disease and pain states by using both unmixing and auto fluorescent features. (A) Clusters of fibres from IA and active fibres, (B) clusters of IA (low and mid) fibres, (C) clusters of active (mid and high) fibres in the two groups, and (D) clustering of IA-Mid and RA-Mid fibres. Symbols represent individual fibres. Ellipses are one standard deviation around the mean, with the central point indicated by a cross.

Figure 7

Bright-field images (first column) of synovial tissues with corresponding images highlighting disease state differences (by CYMK colour, second column). Rows are presented for four different groups: (A) Inactive (IA)-Low, (B) IA-Mid, (C) RA-Mid, and (D) RA-High. Scale bar is representative of 50 µm. Histological assessment of collagen type II abundance within ST was performed to validate the HAI process (E). Error bars represent SEM (IA-Low n = 8, IA-Mid n = 1, RA-Mid n = 4, RA-High n = 9).