Hyperspectral imaging for skin assessment in systemic sclerosis: a pilot study

Abstract

Objective Hyperspectral imaging (HSI) is a novel technology for obtaining quantitative measurements from transcutaneous spatial and spectral information. In patients with SSc, the severity of skin tightness is associated with internal organ involvement. However, clinical assessment using the modified Rodnan skin score is highly variable and there are currently no universal standardized protocols. This study aimed to compare the ability to differentiate between SSc patients and healthy controls using skin scores, ultrasound and HSI. Methods Short-wave infrared light was utilized to detect the spectral angle mapper (SAM) of HSI. In addition, skin severity was evaluated by skin scores, ultrasound to detect dermal thickness and strain elastography. Spearman’s correlation was used for assessing skin scores, strain ratio, thickness and SAM. Comparisons of various assessment tools were performed by receiver operating characteristic curves. Results In total, 31 SSc patients were enrolled. SAM was positively correlated with skin scores and dermal thickness. In SSc patients with normal skin scores, SAM values were still significantly higher than in healthy controls. SAM exhibited the highest area under the curve (AUC: 0.812, P < 0.001) in detecting SSc compared with skin scores (AUC: 0.712, P < 0.001), thickness (AUC: 0.585, P = 0.009) and strain ratio by elastography (AUC: 0.522, P = 0.510). Moreover, the severity of skin tightness was reflected by the incremental changes of waveforms in the spectral diagrams. Conclusion SAM was correlated with skin scores and sufficiently sensitive to detect subclinical disease. HSI can be used as a novel, non-invasive method for assessing skin changes in SSc.

Hyperspectral Imaging Assessment of Systemic Sclerosis Using the Soft Abundance Score and Band Selection

“…In our prior report [21], hyperspectral image analysis demonstrated itself as a feasible and superior method in assessing skin severity in SSc patients. Therefore, we used the SAM method from our previous study as a comparison group [21]. The correlation analysis of the SAM and band selection methods is shown in Table II.…”

“…Ultrasound was performed in both SSc patients and healthy controls using Philips iu22 (Philips Ultrasound, Bothell WA, USA), 4-10 MHz and 15 MHz linear probes to detect dermal thickness as previously described [21]. Six skin regions, namely, the bilateral dorsal hands and the bilateral anterior and upper forearms, were selected as shown in Figure 2.…”

“…In addition, skin scores from 6 skin regions imaged by HSI were also assessed by two independent and experienced rheumatologists [21]. The clinical evaluation used the semiquantitative scoring system of the mRSS [29].…”

“…Dimensionality reduction has two purposes: to reduce the number of calculations for hyperspectral data and to capture spectral bands with essential information before being used as a principle for band selection (BS). In our previous report [21], we did not discuss the influence of band selection methods on the analysis of scleroderma. It remains unknown whether the band selection methods proposed in a prior study [22] could be used to evaluate SSc patients.…”

“…Our group has recently demonstrated the feasibility of applying HSI to assess disease severity in scleroderma patients [21]. Compared with the mRSS, which determines skin thickness by ultrasound and the elastic strain ratio, HSI has a superior diagnostic ability to distinguish SSc patients from healthy controls.…”