Early identification of life-threatening soft-tissue infection using dynamic fluorescence imaging: first-in-kind clinical study of first-pass kinetics

Abstract

Necrotizing soft-tissue infections (NSTIs) are aggressive and deadly. Immediate surgical debridement is standard-of-care, but patients often present with non-specific symptoms, thereby delaying treatment. Because NSTIs cause microvascular thrombosis, we hypothesized that perfusion imaging using indocyanine green (ICG) would show diminished fluorescence signal in NSTI-affected tissues, particularly compared to non-necrotizing, superficial infections. Through a first-in-kind clinical study, we performed first-pass ICG fluorescence perfusion imaging of patients with suspected NSTIs. Early results support our hypothesis that ICG signal voids occur in NSTI-affected tissues and that dynamic contrast-enhanced fluorescence parameters reveal tissue kinetics that may be related to disease progression and extent.

1. INTRODUCTION

Patients with necrotizing soft-tissue infections (NSTIs) have an acute mortality rate near 30%^1–5. An NSTI occurs with an inoculation of virulent bacteria in or around fascia, the connective tissue layer surrounding our muscle compartments. With some bacterial strains (namely, group A streptococcus), this tissue layer provides a relatively immunoprivileged and permissive environment for bacterial growth, facilitating rapid advancement along fascial planes. The result is a soft-tissue infection that quickly spreads centrally—often within hours—leading to sepsis, multi-organ failure, and death. Immediate, aggressive surgical debridement is standard-of-care. A patient with an NSTI presents with non-specific symptoms—fever, pain, and elevated inflammatory laboratory values—and the infection can initiate in deep soft tissues without a clear entry point, such that overlying skin initially appears unremarkable⁴. No definitive diagnostic test exists for NSTIs⁴. Other soft-tissue infections, such as cellulitis, are relatively common and substantially less deadly than NSTIs, but these less serious infections have a similar, non-specific presentation and also lack a definitive diagnostic test⁶. Consequently, correct diagnosis of an NSTI is often delayed, leading to high morbidity³ and mortality⁴.

Fluorescence-guided surgery (FGS) enables the differentiation of key tissues and anatomical structures via the excitation of and emission from fluorescent probes. These probes, known as “fluorophores,” can highlight tissues of interest via local vascular perfusion or via key chemical processes, e.g., molecular binding or metabolism⁷. A key feature of an NSTI is prominent microvascular thrombosis in affected tissues^8,9. We hypothesized that first-pass perfusion fluorescence imaging would show diminished fluorescence signal in NSTI-affected tissues, particularly compared to non-necrotizing, superficial infections (e.g., cellulitis) where local inflammation, edema, and vasodilation occur⁶, hypothetically leading to increased fluorescence signal.

Indocyanine green (ICG) is a near-infrared (NIR) fluorophore (excitation peak range of 780–805 nm, emission peak range of 805–810 nm) with almost 70 years of clinical use¹⁰. ICG has undergone a resurgence of use in recent years primarily for tissue perfusion assessment^7,10,11 but also for tissue identification and soft-tissue infection management¹². Dynamic contrast-enhanced fluorescence imaging (DCE-FI) using ICG involves the acquisition of fluorescence imagery for several seconds to minutes following probe administration. Kinetic parameters extracted from fluorescence time series data can be related to underlying tissue pathophysiology, including the ingress slope (IS), time-to-peak (TTP) intensity, max intensity (Imax), and egress slope (ES) parameters^13,14. The goal of this clinical study is to test the hypothesis that perfusion fluorescence imaging (i.e., ICG-based DCE-FI) shows diminished fluorescence signal in NSTI-affected tissues in patients presenting to the Emergency Department (ED) for soft-tissue infection treatment.

2. METHODS

2.1. Study Design

The study design is summarized in Figure 1. A first-in-kind clinical study approved by the Dartmouth Health Institutional Review Board was initiated to test our hypothesis that NSTI-positive tissues exhibit diminished perfusion fluorescence (ClinicalTrials.gov Identifier: NCT04839302). Patients presenting to the Emergency Department at Dartmouth-Hitchcock Medical Center (DHMC) in Lebanon, New Hampshire, with clinical features of an NSTI and a Laboratory Risk Indicator for Necrotizing Fasciitis score ≥6^2,5 were promptly referred by ED providers to the study team. The study team educated the patient about the study, and each consenting patient was imaged using the SPY Elite imaging system (Stryker-Novadaq, Kalamazoo, MI). Clinicians providing standard-of-care treatment were blinded to all imaging performed in this study. Imaging occurred at video rate for 10 seconds prior to intravenous administration of a weight-based dose of ICG (0.1–0.2 mg/kg) to establish a baseline and proceeded for four minutes post-injection to capture fluorescence time series data. One or more white light images were also acquired. Imaging encompassed the affected site along with adjacent, unaffected tissues for intra-patient control. After imaging, patients proceeded with standard-of-care management, involving surgical debridement in the operating room (OR), antibiotic treatment, and confirmation of NSTI status based on bacterial culture and histopathologic diagnosis of tissue specimens. The study protocol is approved for enrollment of 40 patients total.

Figure 1.

Study design flowchart. ED = Emergency Department. ICG = indocyanine green.

2.2. Image Analysis

All image analysis was performed using MATLAB (version R2022a, MathWorks, Natick, MA), the Image Processing Toolbox (version 11.5), and the Statistics and Machine Learning Toolbox (version 12.3). An in-house script performed ICG intensity curve parameterization, generating wide-field maps of DCE-FI kinetic parameters (i.e., IS, TTP, Imax, and ES)^13,14. A maximum of 1024 frames were acquired by the SPY Elite system per patient. A representative perfusion fluorescence intensity curve is shown in Figure 2 along with curve parameters of IS, TTP, Imax, and ES.

Figure 2.

Illustration of ICG fluorescence intensity curve (green) parameterization, yielding four parameters: ingress slope (IS), time-to-peak (TTP), max intensity (Imax), and egress slope (ES). Figure adapted from Streeter et al.¹².

Regions of interest (ROIs) were selected from each fluorescence time series image stack. In necrotizing fasciitis cases, the ROIs highlighted tissues suspected to be NSTI-positive, borderline tissues near the edge of affected and unaffected tissue regions, and tissues suspected to be unaffected by the infection. ICG fluorescence intensity curves from the ROIs were visualized. Principal component analysis (PCA, implemented via the “pca” function with default settings) was performed using a single patient’s data to demonstrate ROI separability based on DCE-FI kinetic parameters. Histopathologic diagnosis of tissue specimens provided ground truth for NSTI status.

3. RESULTS

Results from the first eight patients enrolled in the clinical study are summarized in Table 1. To date, no adverse events have occurred due to intravenous administration of ICG, and no delays to standard-of-care treatment have been reported. Of the first eight enrolled, one patient was not imaged (Patient 03), because an OR became available before fluorescence imaging could be performed, and it was not in the best interest of the patient to delay standard-of-care treatment. The remaining seven patients had confirmed cases of necrotizing fasciitis (Patients 02, 04, and 07), cellulitis (Patients 01, 05, and 06), or gangrene (Patient 08). None of the imaged cellulitis infections later progressed to an NSTI. Necrotizing fasciitis and cellulitis cases are summarized in Figure 3. All cases of necrotizing fasciitis had prominent first-pass fluorescence signal voids suspected to be NSTI-positive tissue regions (regions delineated by yellow dashed lines in Figure 3a–Figure 3c), whereas all cases of cellulitis lacked a definitive signal void.

Table 1.

Enrolled patient demographics and imaging results.

Patient ID	Sex	Age	Infection Location	Final Diagnosis	Imaging Performed	Prominent Signal Void
01	F	75	Quadricep, left	Cellulitis	Yes	No
02	F	59	Lower limb, left	Necrotizing fasciitis	Yes	Yes
03	--	--	--	--	No	--
04	F	78	Quadricep, left	Necrotizing fasciitis	Yes	Yes
05	M	36	Lower limb, left	Cellulitis	Yes	No
06	M	70	Lower limb, left	Necrotizing fasciitis	Yes	Yes
07	F	62	Lower limb, right	Cellulitis	Yes	No
08	M	48	Big toe, left	Gangrene	Yes	Yes

Figure 3.

Summary of preliminary (a-c) pathology confirmed necrotizing fasciitis cases and (d-f) cellulitis cases imaged in the clinical study. White light images of affected tissues are shown alongside representative ICG fluorescence images of the same tissue. Yellow dashed lines in (a-c) delineate suspected regions of infection. A fluorescence imaging standard developed by QUEL Imaging (White River Junction, VT)^15,16 is visible in (b-c) and (e-f).

ROI-derived fluorescence intensity curves and example parameter maps are shown for the three necrotizing fasciitis cases in Figure 4 and for the three cellulitis cases in Figure 5. In the necrotizing fasciitis cases, suspected regions of affected tissue (red ROIs) exhibited max intensities ≤120 a.u. and egress slopes ≤0.6 a.u. while suspected regions of unaffected tissue (blue ROIs) exhibited max intensities of ≥220 a.u. and egress slopes of ~1.0 a.u.

Figure 4.

Regions of interest in three confirmed necrotizing fasciitis cases (shown in column 1) delineate tissues suspected to be NSTI-positive (red), borderline tissues (green), and tissues suspected to be NSTI-negative (blue). ICG fluorescence intensity time profiles (column 2, colored interval = ROI pixel intensity average ± one standard deviation) enabled quantification of dynamic contrast-enhanced fluorescence parameter maps, including maps for ingress slope (column 3) and max intensity (column 4). The pink arrow in (b) column 2 highlights a secondary peak in intensity due to ICG recirculation. Red arrows in (c) column 2 highlight motion artifacts. Patient 04 images adapted from Streeter et al.¹²

Figure 5.

Regions of interest (shown in column 1) in three confirmed cellulitis cases positioned uniformly across each field of view. Corresponding ICG fluorescence intensity time profiles (column 2, colored interval = ROI pixel intensity average ± one standard deviation) enabled quantification of dynamic contrast-enhanced fluorescence parameters, including maps of ingress slope (column 3) and max intensity (column 4). The pink arrow in (c) column 2 highlights a secondary peak in intensity due to ICG recirculation. A fluorescence imaging standard developed by QUEL Imaging (White River Junction, VT)^15,16 is visible in (b) and (c).

The case of gangrene (Patient 08) presented with a prominent fluorescence signal void in their affected toe, similar in appearance to the necrotizing fasciitis cases (data not shown). However, the patient reported symptoms for over one week prior to presenting to the DHMC ED. Being symptomatic for several days is uncharacteristic of an NSTI, which typically progresses in severity rapidly, over the course of hours. Thus, clinical context could have potentially ruled out Patient 08 for an NSTI.

To demonstrate the potential value of DCE-FI kinetic parameters for differentiating tissue regions, the necrotizing fasciitis case with the most complete and artifact-free ICG fluorescence time series data (Patient 04) was further analyzed. The completeness of the Patient 04 time series data enabled the quantification of IS, TTP, Imax, and ES DCE-FI kinetic parameters over the entire field of view (Figure 6a–Figure 6b). PCA was applied to only Patient 04 data encompassed in the three ROIs; individual pixels were considered observations, while all four DCE-FI kinetic parameters were considered variables. The first three principal components showed clear separation of pixels contained in each ROI (Figure 6c).

Figure 6.

Demonstration of principal component analysis using Patient 04 data. (a) Patient 04 ROIs depict suspected affected tissues (red), borderline tissues (green), and suspected normal tissues (blue). (b) Dynamic contrast-enhanced fluorescence parameter maps for ingress slope, time-to-peak, max intensity, and egress slope. (c) Principal component analysis using ROI pixels as observations and parameter values at each ROI pixel as variables. Patient 04 fluorescence and parameter map images adapted from Streeter et al.¹²

4. DISCUSSION

Preliminary results support the hypothesis that ICG-based first-pass DCE-FI shows diminished fluorescence signal in NSTI-positive tissues. Importantly, these fluorescence signal voids do not always align with features seen in the white light imagery, demonstrating how the technique might improve the surgeon’s ability to identify affected tissues. Furthermore, DCE-FI kinetic parameters may provide value for differentiating NSTI-positive, borderline, and NSTInegative tissues beyond that provided by single snapshot fluorescence imaging. DCE-FI took approximately five minutes per patient, and thus, could act as a rapid means of triaging NSTI-positive cases upon initial presentation for medical treatment.

Limitations of the data collected thus far include incomplete ICG fluorescence intensity time series (e.g., Figure 4a, Figure 5a), time series artifacts incurred by patient and/or imaging system movement (e.g., red arrows in Figure 4c), saturation of the detector due to imprecise ICG dosing and/or image acquisition settings (e.g., Figure 4a–Figure 4b, Figure 5a–Figure 5c), and limited detector dynamic range (the SPY Elite is capable of only eight-bit image acquisition). A key limitation of our study is that only patient-level histopathological diagnoses are currently provided based on standard-of-care bacterial cultures. Thus, image features (e.g., ICG fluorescence signal voids) are suspected to be associated with (un)affected tissues, but spatially resolved histopathology is unavailable to confirm the relationship between fluorescence signatures and disease states. A final noteworthy limitation is the effects of tissue surface curvature, which impact measured fluorescence intensity and downstream DCE-FI parameterization (as seen along the edges of all imaged cases in Figure 4 and Figure 5); three-dimensional/depth sensing imaging technologies could in theory be used correct for these effects.

Ongoing work includes further analysis of ICG fluorescence curve features and how they relate to underlying tissue pathophysiology. Improved imaging instrumentation is also actively being pursued through the development of an in-house ICG fluorescence imaging system capable of greater dynamic range (10-bit) image acquisition.

Future work involves collecting punch biopsies for spatially resolved histopathological confirmation of infection that can be co-registered with ICG fluorescence signatures. Steps to make ICG-based DCE-FI more quantitative and generalizable between patients and pathologies will also be pursued, including the integration of a patient-specific arterial input function to the parameterization process¹³ and the use of a fluorescence imaging standard in the field of view to enable repeatable measurements of fluorescence intensity. An example fluorescence standard developed by QUEL Imaging (White River Junction, VT)^15,16 is shown in Figure 3–Figure 5 with an ICG fluorescence intensity profile demonstrated by the yellow ROI/curve in Figure 5b. The depth sensitivity of ICG-based DCE-FI for NSTIs depends on patient-specific skin tissue optical properties, local vasculature, and the size and location of the infection, but in general, is estimated to be millimeters. Techniques to quantify the depth of fluorescence emission could be coupled with DCE-FI, such as fluorescence LiDAR¹⁷, to potentially detect infection-related fluorescence signatures at depth. Finally, future work will involve the development of efficient, streamlined analysis of image data such that clinicians are provided with a “NSTI-positive/NSTI-negative” interpretation of target tissues, mitigating the need for providers to do their own image interpretation.

Patient enrollment and imaging are ongoing. Once additional patients with suspected NSTIs have been imaged and robust ICG-based DCE-FI signature/histopathological correlates have been determined, a prospective, multi-center observational trial of ICG fluorescence in patients presenting to tertiary EDs with possible NSTIs will be pursued. ICG will be administered with the goals of, first, accurately distinguishing NSTIs and non-necrotizing soft-tissue infections, and second, correlating the use of ICG fluorescence with surgical debridement success.