Is my wound infected? A study on the use of hyperspectral imaging to assess wound infection

Abstract

Introduction: Clinical signs and symptoms (CSS) of infection are a standard part of wound care, yet they can have low specificity and sensitivity, which can further vary due to clinician knowledge, experience, and education. Wound photography is becoming more widely adopted to support wound care. Thermography has been studied in the medical literature to assess signs of perfusion and inflammation for decades. Bacterial fluorescence has recently emerged as a valuable tool to detect a high bacterial load within wounds. Combining these modalities offers a potential objective screening tool for wound infection.

Methods: A multi-center prospective study of 66 outpatient wound care patients used hyperspectral imaging to collect visible light, thermography, and bacterial fluorescence images. Wounds were assessed and screened using the International Wound Infection Institute (IWII) checklist for CSS of infection. Principal component analysis was performed on the images to identify wounds presenting as infected, inflamed, or non-infected.

Results: The model could accurately predict all three wound classes (infected, inflamed, and non-infected) with an accuracy of 74%. They performed best on infected wounds (100% sensitivity and 91% specificity) compared to non-inflamed (sensitivity 94%, specificity 70%) and inflamed wounds (85% sensitivity, 77% specificity).

Discussion: Combining multiple imaging modalities enables the application of models to improve wound assessment. Infection detection by CSS is vulnerable to subjective interpretation and variability based on clinicians' education and skills. Enabling clinicians to use point-of-care hyperspectral imaging may allow earlier infection detection and intervention, possibly preventing delays in wound healing and minimizing adverse events.

Keywords: bacteria; fluorescence; hyperspectral imaging; infection; inflammation; point-of-care; thermography; wounds.

Figure 1

The Ray 1 hyperspectral imaging device. The Ray 1 imaging device is a pocket-sized hyperspectral camera that is designed to sit over a smartphone's camera lens and wirelessly connect to Swift Medical's Skin and Wound app. Once connected, the camera allows the simultaneous acquisition of: visible light images that can be used for clinical inspection, wound area measurement, and the automated identification of tissue types present in the wound; infrared thermal images for the assessment of vascular and inflammatory patterns; and bacterial fluorescence images for the assessment of bacterial bioload in wounds, such as in the vignette presented at the bottom of the figure.

Figure 2

Hyperspectral imaging patterns. The assessment of hyperspectral images of wounds revealed four distinct imaging patterns. The first one corresponds to the non-inflamed wounds characterized by colder wound and peri-wound areas and negative bacterial fluorescence. The second one corresponds to the inflamed wounds characterized by hotter peri-wound areas and cold to moderately warm wound beds with negative to slightly positive bacterial fluorescence. Finally, the third and fourth patterns correspond to the infected wounds characterized by frank hotspots compatible with areas of extensive inflammation with or without positive bacterial fluorescence.

Figure 3

Thermal and bacterial fluorescence quantitative assessment. Quantitative assessment of the infrared thermographic images (A) showing a gradient of the severity of the thermal asymmetry of the peri-wound area compared to adjacent healthy skin between the non-inflamed, inflamed, and infected wounds. In contrast, for the quantification of the area positive for bacterial fluorescence (B), the non-inflamed and inflamed wounds do not exhibit a clear gradient, while the infected wounds show a spectrum of positive fluorescence ranging from none detected to over 10 cm² of fluorescence area.

Figure 4

PCA-KNN clustering. Data dimensionality reduction was performed through a principal component analysis (PCA), including clinical and hyperspectral imaging data. K-nearest neighbor (KNN) clustering of the two principal components allows the classification of data points as belonging to three distinct classes encompassing the non-inflamed, inflamed, or infected wounds. The accuracy of the model was 74%. The large circles inside the clusters represent their centroids.