多光谱图像
光谱成像
计算机视觉
计算机科学
考试(生物学)
医学物理学
人工智能
遥感
医学
地质学
古生物学
摘要
Lisa Kiely,1 Michael McAuliffe,2 Aoife Hollywood1 and John F. Bourke1 1South Infirmary Victoria University Hospital and 2Munster Technological University, Cork, Ireland The COVID-19 pandemic led to the implementation of virtual patch testing in order to maintain access to services. Studies examining this practice, using standard digital cameras to identify positive reactions, were found to be inferior to clinical assessment. Digital cameras have traditionally come in two formats: monochrome and colour. Multispectral cameras target spectral bands in both visible and nonvisible spectral regions to perform more sophisticated analysis. Using this technology, it is possible to assess levels of blood flow in a clinical photograph of skin acting as a surrogate marker for cutaneous inflammation. Multispectral imaging has never been used for patch test analysis. The aim was to assess the spectral response between 400 and 1000 nm of positive reactions of patch test results. Patients attending for patch testing were recruited and written consent was obtained. Patients returned on day 5 for clinical assessment. Digital photographs were taken and results were determined to be positive, irritant or other. Spectra were recorded using an Ocean optics QE pro spectrometer and a tungsten halogen source, both coupled with a bifurcated optical fibre. The fibre had a spacer so that a fixed optical distance could be maintain upon contact with the patient. Spectra were recorded in triplicates of positive reactions, irritants and a control negative result. Reference spectra were also collected away from the patch test area as an additional control. The spectra were analysed using principal component analysis (PCA), a technique for reducing the dimensionality of datasets, increasing interpretability while minimizing information loss. Spectra were treated using a combination of different mathematical transformation. The results were then examined to determine the optimal wavelengths required for differentiating between positive, irritant and control results. Nine consecutive patients were recruited from September to November 2021. Spectra from 40 positive reactions, two dyes and one irritant were recorded. Based on the spectra collected and analysed the optimum wavelengths were found between 510 nm and 620 nm. PCA results provide an indication of the severity of the reaction based on the clinical diagnosis. It was also possible to differentiate between dye results and positive results. This pilot study outlines the potential application of multispectral imaging for the interpretation of patch test results. The results show there is a spectral difference between positive and negative results, which would support the development of a multispectral camera that could support clinical diagnoses of virtual and remote patch testing.
科研通智能强力驱动
Strongly Powered by AbleSci AI