Fluorescence-guided and molecularly-guided debridement: identifying devitalized and infected tissue in orthopaedic trauma

Abstract

Following orthopaedic trauma, bone devitalization is a critical determinant of complications such as infection or nonunion. Intraoperative assessment of bone perfusion has thus far been limited. Furthermore, treatment failure for infected fractures is unreasonably high, owing to the propensity of biofilm to form and become entrenched in poorly vascularized bone. Fluorescence-guided surgery and molecularly-guided surgery could be used to evaluate the viability of bone and soft tissue and detect the presence of planktonic and biofilm-forming bacteria. This proceedings paper discusses the motivation behind developing this technology and our most recent preclinical and clinical results.

1. INTRODUCTION

1.1. Clinical Problem

Orthopaedic injury is considered one of the most common reasons for a hospital visit, comprising 10% of visits to the emergency department and about 5% of all hospital admissions.¹ Many orthopaedic injuries are caused by low-energy mechanisms such as certain types of falls or assaults and overexertion. Injuries caused by high-energy mechanisms such as firearms, motor vehicles, explosions or farm/industrial machinery have a high-risk of complications. One such complication is the development of infection. The presence of metallic implants needed to stabilize the fracture, along with contamination to the wound during the initial injury, provides a challenging combination of factors that make infection very difficult to treat once it has established.

Infection requires one or more unplanned surgical procedures and leads to prolonged morbidity, loss of function, and potential loss of limb.^2–11 About 60% of high-energy open fractures result in the development of infection. For these patients, initial treatment of the infection fails 30% of the time, and the consequences are profound. Many of the late amputations performed among military service members, for example, are due to persistent deep infection and/or osteomyelitis.¹²

Management of implant-associated infection is both surgical and medical, and includes thorough surgical debridement of infected and/or poorly perfused bone and soft tissue, removal of metallic implants, mechanical stability of the extremity, soft tissue coverage and a combination of local and systemic antibiotics.¹³ Because of the propensity for biofilm to recur on bone and newly implanted hardware, fracture-related infections are often also staged using external fixation devices and/or temporary antibiotic-coated implants. These staged procedures represent a high burden for patients and an extraordinarily high economic cost.

The goal of surgical debridement for infection is to remove all bone and soft tissue that cannot be effectively treated using antibiotics and that can form a nidus for infection. Antibiotics cannot penetrate poorly perfused tissues. However, perhaps more importantly, established bacterial infections evade treatment through the formation of biofilms—a growth mode in which microbial communities are spatially structured and embedded in a matrix of extracellular polymeric substances (EPS)¹⁴. Biofilm forms a mechanical barrier to protect against treatment and also acquires pathogenicity and antimicrobial resistance through their ability to exchange information with each other through quorum sensing^15–17. Studies have demonstrated that the minimum inhibitory concentration (MIC) of an antibiotic to inhibit bacterial growth of biofilm-based bacteria is up to 1000x higher than planktonic or free-floating bacteria, which is far above the safe dosage range for human use.¹⁸

Deficient perfusion prevents delivery of antibiotics and endogenous immune cells to affected bone and soft tissue.³ Poorly perfused bone becomes a nidus for biofilm formation¹⁹, effectively behaving as a foreign body, rendering antibiotics (and endogenous immune cells) ineffective.

Additionally, many microbial communities can display a broad range of behaviors that make them resistant to antibiotic treatment independent of the vascular perfusion of the bone or tissue, such as secreted toxins, immune evasion²⁰, intracellular persistence^21,22, biofilm formation^23,24, antimicrobial resistance genes, the creation of slow growing small colony variant subpopulations^25,26, the development of antibiotic resistance²⁷, and invasion of lacuna-canalicular network of live bone^28,29.

It is becoming increasingly clear that microbes can take advantage of unique features bone anatomy such as the small lacuna-canalicular networks. Even in viable, well perfused bone, microbes can hide in these small networks which are impenetrable by either antimicrobials or endogenous immune cells can penetrate. Furthermore, biofilm may more easily form on bone due to the unusual vascular anatomy. It is clear that metallic implants and other foreign bodies are inherently vulnerable to bacterial contamination and biofilm formation³⁰ and that microbial communities are extraordinarily difficult to remove from these surfaces, once they become established.

If antibiotics are ineffective at eradicating remaining microbial communities characterized by (a) lack of access (inadequate vascular perfusion), (b) microbial behaviors such as biofilm production, (c) properties of bone and/or (d) presence of hardware, the infection will persist requiring additional procedures with worse function, health-related quality of life and a high rate of conversion to amputation^{12,13,31–33}.

Therefore, success of targeted antimicrobial therapy depends on complete surgical debridement of both devitalized and biofilm-containing tissue. The optimization of debridement is critical: insufficient debridement will place patients at high risk for treatment failure and recurrent infection, with severe consequences. However, excessive debridement removes healthy tissue, increasing the complexity of reconstructive procedures with inferior functional outcomes. Unfortunately, there are no methods to precisely identify tissues that must be debrided. This reality is undoubtedly a contributing factor to poor cure rate of upwards of 30% in patients who developed recurrent or persistent infection following treatment of an implant-associated infection, even using the two-stage approach. An effective intraoperative method to precisely guide debridement will direct the surgeons’ attention to two critical classes of tissue: (1) devitalized or poorly perfused bone and soft tissue; and/or (2) infected bone and soft tissue. There is clearly overlap between bone/tissue devitalization and bone/tissue infectious infiltration in that devitalized tissues are more likely to be infiltrated with invading microbes. However, this is not believed to be a one-to-one relationship because microbial infiltration can extend beyond the devitalized areas, as is seen in the GI system and the oral cavity where well-perfused tissues remain covered with biofilm.

Dynamic contrast-enhanced fluorescence imaging has been proposed as a method of evaluating the viability of tissue and degree with which it has been damaged by the initial injury. This technique, also called laser-assisted angiography in some literature, involves the introduction of fluorescent contrast agent (namely, indocyanine green [ICG]) into the patient intravenously, followed by imaging the fluorescence enhancement of tissue at the bandwidth corresponding to the emission peak. The intensity images recorded using the imaging system can be viewed in real time, or can be collected as a series, and used to calculate perfusion-related parameters such as maximum, time-to-peak, ingress, and egress slope, as well as (with the use of certain more sophisticated models) blood flow, blood volume, extraction fraction, rate of efflux from vasculature to tissue compartment, etc. With the addition of targeted fluorophores, molecular imaging can be performed. When done intraoperatively, this is called “molecularly-guided surgery” or “live surgical navigation.” Regions-of-interest can be quantified according to perfusion and molecular imaging parameters, and can also be analyzed using spatiotemporal or texture analysis. This spatial pattern recognition can yield additional insight into the underlying mechanics and viability.

2. MATERIALS & METHODS

2.1. Fluorescence Imaging Systems

Preclinical imaging

The imaging system used in the acquisition of rat preclinical data is a modified operating microscope equipped with a scientific CMOS camera (Panda 4.1, PCO Tech, Germany) integrated into the Zeiss FC-1 platform. To the bottom of the microscope head is affixed a 760 nm high-powered LED (Mightex Systems, North York, Ontario, Canada), coupled with a short-pass filter, which provides excitation light during imaging. The emission is collected by the objective of the microscope, magnified between 0.6x and 2.4x, and then directed towards a dichroic mirror. The dichroic mirror divides light above 770 nm towards the sCMOS camera, where it is detected just after being filtered with a long-pass filter (780lp, Chroma).

Clinical imaging

In the clinical investigation, the SPY Elite system (Stryker Corp., Kalamazoo, MI, USA) was used for the majority of the data acquisitions. We also use the SPY PHI system (Stryker Corp.) and have developed an adapter that can provide working distance calibration and simultaneous white-light video. All studies were approved by the local Institutional Review Board (IRB) at Dartmouth Health, which also served as the central IRB for the multisite study; the studies are listed on clinicaltrials.gov (NCT04245111, NCT04250558, NCT04403204 and NCT04416412). In all cases, informed consent was obtained prior to surgery. For 20 seconds before and for 4 minutes after the injection, fluorescence images of surgical areas were recorded using SPY Elite equipped with 805 ± 10 nm LED for ICG excitation and NIR charge-coupled device camera with 820 – 900 nm band pass filter. Recorded images were processed by in house developed software (MATLAB R2022a, MathWorks, Natick, MA), which performed image co-registration, motion correction, simple fitting to extract maximum intensity (I_max), ingress (IS), egress slope (ES) and time-to-peak (TTP), as well as kinetic modeling.

2.2. Tracer Kinetics

Tracer kinetic theory can be used to describe the behavior of a bolus (i.e., an approximate Dirac-delta function) as it is distributed by the vasculature and transferred between tissue compartments in a tissue region of interest. In the general case, this time-dependent concentration, C(t), measured in the tissue can be described as:

$$C(t)=\text{F}R(t)\ast C_a(t)$$

where F is the blood flow, C_a(t) is the arterial input function and R(t) is the impulse response function—the fraction of ICG remaining in the tissue at time, t, for the idealized case that the input function is defined by a Dirac-delta function with unit mass deposited at time zero. Since C_a(t) is often a bolus injection with a finite width that has been recirculated one or more times before it reaches the tissue of interest, it will have an appearance like a sharp peak with smaller bumps 30 seconds or so after. This function is modified by the unique hemodynamic properties of the tissue of interest, and results in the C(t) measured by the DCE-FI system.

The simplest way to approach the analysis of DCE-FI data is to ignore between-patient differences in AIF, and directly characterize the C(t) function by extracting simple parameters: maximum intensity, ingress slope, egress slope, and time-to-peak. However, as we demonstrated in our 2020 paper showing the effect of AIF perturbation on simple variables,³⁴ this approach is corrupted by between-subject variability in AIF shape and intensity, which would be very much expected in an application such as this, where polytrauma and systemic instability is common, and where conditions like cardiovascular health and diabetes can effect outcomes. Our proposed correction method deconvolves the C(t) and then reconvolves it with a standardized AIF, to produce C_corr(t), or corrected tissue curve.

Kinetic modeling is more computationally demanding and requires some a priori knowledge of the underlying tissue behavior and composition, but can recover values that might be more physiologically meaningful. For this, we apply three different models: (1) the St Lawrence & Lee model (also called the adiabatic approximation to the tissue homogeneity model [AATH])^35–37, (2) the gamma capillary transit time (GCTT) model³⁸ and (3) the hybrid plug-compartment (HyPC) model³⁹. These in their R(t) functions, when R(t) is defined generally R₁(t) + R₂(t), and the explicit models are provided in Table 1.

Table 1.

Impulse residue functions for the three models used in this paper.

	R1(t)	R2(t)	Free Parameters
AATH	1 − Θ(t − tc)	$E{e}^{-k_{ep}\left(t-t_c\right)}\Theta \left(t-t_c\right)$	4
DCTT	$1-{\int }_0^1𝒟(u)du$	$E{e}^{-k_{ep}t}{\int }_0^1𝒟(u)e^{-k_{ep}(t-u)}du$	5
HyPC	$\{\begin{array}{ll}\hfill 0\hfill & t<T_E\hfill \\ \hfill \raisebox{1ex}{$EBF$}\!\left/ \!\raisebox{-1ex}{$F$}\right.\hfill & T_E<t<T_E+M_E\hfill \\ \hfill 0\hfill & t>T_E+M_E\hfill \end{array}$	$\{\begin{array}{cc}0& t<T_L\\ \raisebox{1ex}{$LBF$}\!\left/ \!\raisebox{-1ex}{$F$}\right.e^{-k_2\left(t-T_L\right)}& t>T_L\end{array}$	6

T_c – capillary transit time, E – extraction fraction, k_ep, k₂ –extravascular to intravascular rate constant, EBF – early blood flow, LBF – late blood flow, F – (total) blood flow, T_E – appearance time for early phase vasculature, T_L – appearance time for late phase vasculature, M_E – minimum transit time across early phase vasculature.

2.3. Nonparametric approaches

Nonparametric analysis involves extracting R(t) without explicit modeling or parameterization of it. The benefit of this approach is that you recover something that is free to lay itself out in a manner reflective of the ‘black box’ properties, and not shoehorned into a predefined model. This can be particularly beneficial for more complex vascular systems whose underpinning structure is not easily known. However, deconvolution can be unstable. We have explored deconvolution using the truncated singular value decomposition (SVD) approach which imposes no physiological constraints. This is done in MATLAB by singular value decomposition of the Toeplitz matrix of C_a(t) and computing the pseudoinverse of that matrix truncated to 11 singular values with the vector form of C(t).^40,41 Following recovery of FR(t), statistical moments can be used to extract feature maps (e.g., median, area under the curve, variance, skewness, etc).

3. PRECLINICAL MODEL

3.1. High-Energy Contaminated Fracture Model

The complete details of the blast overpressure tube system are reported elsewhere.⁴² Breifly, the blast overpressure tube system was fabricated in-house from a polyvinyl chloride Schedule 40 (max working pressure = 1930 kPa) cylindrical tube (Charlotte Pipe, Charlotte, NC), separated using flanges into 2-ft long compression and 6-ft long expansion chambers by 3 to 5 mil (76 to 127 μm) thick polyethylene terephthalate (Mylar) membrane. An air compressor was connected to the compression chamber through a flexible hose and open-relief valve assembly. The air compressor pressure regulator was set to 145 psi (1000 kPa) with a relief valve protecting the system from overpressure. By opening the valve, the compression chamber was continuously filled with air until the Mylar membrane spontaneously ruptured, causing a shock wave to propagate through the expansion chamber exiting through the reducer and impacting the object. A custom animal holder was designed, 3D-printed using polylactic acid (Pruza Polymers, Prague, Czech Republic) and fixed on an adjustable platform mounted on aluminum breadboards (MB1212 and MB2424, Thorlabs, Newton, NJ).

This trauma model is capable of producing Grade I-IIIc injuries, which are appropriate for validating imaging methods against a gold standard (e.g., fluorescence microspheres or DCE-uCT), for determining the effectiveness of fluorescence-guided debridement and molecularly-guided debridement with or without adjuvant photodynamic therapy treatment (figure 4).

Figure 4.

(A) White-light image showing the tibia fracture and open wound caused by the overpressure tube. The fracture site and surrounding soft tissue was inoculated with MRSA strain SAP231, which had been cultured overnight to log-growth phase. (B) After 1 hour of incubation with 5-ALA, there was a marked reduction in bioluminescent signal (C, before treatment and D, post treatment).

Ongoing studies are evaluating the following: effectiveness of debridement under fluorescence image guidance vs white-light only; association between infection-free survival (or time-to-event) and treatment with biofilm-targeting antibodies; effectiveness of antimicrobial photodynamic therapy (aPDT) on preventing infection (fig. 4.).

4. CLINICAL STUDIES

Clinical studies are ongoing at multiple sites and under three active NCT registrations. To date, about 50% of the patients have been enrolled.

Detailed results from each of these studies are forthcoming, but for the scope of this proceedings, an example of data acquired from a tibia (Figure 6). The map on the left shows the region corresponding to the tibia (dashed line). The region in the middle, just distal to the fracture site, shows substantial decrease in intensity. Examining the time-concentration curve of corresponding to three ROIs, you see markedly different wash-in and wash-out kinetics.

Figure 6.

(A) The maximum fluorescence intensity corresponding to tibia. ROIs selected along the length of the bone show vastly different kinetics. It is this shape that contains information in addition to intensity, such as time-to-peak, and wash-in and wash-out slope.

This can be performed for each subject, as described above and correcting for effects of cardiovascular health by the arterial input function. This was done for all subjects in the “Infection” cohort. There were 28 subjects for which follow up data up to twelve months was available. When considering patients who became infected by 6 months vs patients who healed, the average maximum fluorescence in the bone ROI associated with fracture damage was significantly higher for those patients who healed vs those who had infection or complications (such as conversion to amputation). When considering two groups of patients: those whose maximum fluorescence value is less than 60 rfus vs those with higher fluorescence, the outcomes of these subjects showed different progression. These two observations indicate the importance of perfusion in whether subsequent infection will occur.

5. DISCUSSION

Fluorescence-guided debridement and fluorescence-guided surgery are very promising applications of intraoperative fluorescence imaging. As the techniques are refined, made more quantitative and comparable across subjects, and more accessible to surgeons we suspect that the impact of this field will continue to grow. Currently, commercial systems have focused on a handful of applications such as oncology, gastrointestinal and cerebrovascular. However, orthopaedic trauma is one of the most common surgeries and reasons for hospital admission, and likely represents a market at least 10x greater.

We show in this paper—along with other papers from our team in this year’s proceedings—including papers 12358–12, 12361–37, 12361–26, 12361–27 and 12361–15—that ICG can be used to distinguish between diseased and healthy tissue in the context of orthopaedic injury and soft-tissue trauma. Thus it could be used to guide surgical practice, showing surgeons the most important or concerning areas that need to be removed. Our hypothesis is that the most important impact of fluorescence-guided and molecularly-guided debridement is to improve the performance of non-experts such as junior residents and attendings, surgeons operating in more rural hospitals, and surgeons operating in austere or constrained environments.

Currently, in most developed countries, the hospital system is organized into local or rural hospitals, regional hospitals, larger hospitals with Level I/II trauma centers and highly specialized care. Natural disasters as well as widespread armed conflict results in significantly higher orthopaedic trauma prevalence as well as increased demands placed on local hospitals to provide trauma care. For example, the invasion of Ukraine in 2022 by Russian forces led to significant military and civilian casualties and destruction of infrastructure. Extensive use of high kinetic energy weapons typical of conventional warfare has led to a very high prevalence of limb injuries, and an increase in the severity of these injuries.

The National Military Medical Clinical Center (NMMCC) in Kyiv reports that 89% of combat injuries are extremity wounds, and 36% of these involve fractures. Based on recent casualty figures, it can be estimated that at least 20,000 Ukrainian soldiers have sustained combat-associated extremity fractures since the start of 2022, though exact figures are classified. Ukraine had configured their evacuation and medical care in accordance with NATO standards (i.e., role 1: first aid; role 2: mobile hospital; role 3: specialized hospital; role 4: evacuation and definitive care in Western Ukraine). However, full-scale war has necessitated flexibility in this system, especially in the face of loss of infrastructure and dramatic increase in casualties. Civilian hospitals have assumed the role of mobile military hospitals (Role 2), performing definitive orthopaedic trauma treatment. The relative inexperience of these small hospitals have resulted in inadequate primary surgical treatment, incorrect timing of internal fixation, and increased contamination of surgical wounds.⁴³

Other situations necessitate that difficult surgical tasks are completed by less experienced practitioners, such as natural disaster, constraints of rural healthcare delivery, and simply the need for less experienced surgeons to gain expertise. To date, we have demonstrated that ICG-based DCE-FI has considerable promise in providing surgeons with an objective measure of bone and soft tissue viability. As a result, DCE-FI could significantly enhance the management of severe traumatic injury. By providing surgeons with quantitative objective data on vascular perfusion of surgical bone and soft tissues, we believe improved ability to assess and perform effective debridement will result. This, in turn, would likely result in decreased variability in treatment and possibly reduced the rate of complication.

Figure 1.

Hypothesized two-variable clustering of values categorized as healthy tissue, devitalized tissue, and infected tissue. An untargeted dye, such as indocyanine green, could be combined with an infection-targeted dye, to discriminate between these three categories, since the one of the disease-states overlaps considerably with normal tissue when considering ICG only.

Figure 2.

Fluorescence imaging system that has been prototyped for remote or austere environments.

Figure 3.

Recovery of FR(t) using nonparametric methods. Deconvolution with a regularization method like truncated singular value decomposition (a) can result in artifacts like ripples. Physiological constraints can stabilize the process, but might not be valid for bone and its overlapping by non-communicative periosteal and endosteal vascular system with separate inputs.

Figure 5.

Enrollment numbers for three clinical studies funded by DOD and NIH grants.

Figure 7.

Interim results from the Infection Study. Twenty-six patients were imaged following debridement to clear infection. (A) The maximum fluorescence within the ROI identified by the surgeon as the most concerning location following debridement, for patients who had healed from infection after 6 months, vs. patients who became infected by 6 months. (B) The maximum fluorescence was used to divide patients into two groups: those with Imax > 60 r.f.u. and those with Imax < 60. The proportion of patients without negative outcome (positive infection, explantation of hardware, or conversion to amputation) was determined from follow up data.