Diagnostic Accuracy of a Fluorescence Imaging Device in Diabetic Wounds: A Pilot Study Using a Tissue Culture System

DFUs are challenging chronic wounds due to their vulnerability to many factors that impair the wound healing process.¹ Although several factors are involved in delayed or nonhealing foot ulcers in patients with diabetes, infection is the most direct and important factor that influences wound healing. Even with proper management, diabetic foot infections can be difficult to control and may lead to amputation. Normally, wound infections manifest as local (eg, pain, erythema, edema, and warmth) or systemic (eg, fever and elevated white blood cell count) symptoms, depending on the severity of infection. However, these symptoms and signs may not be distinct in patients with DFUs because of their poor ability to detect bacterial invasion and diminished inflammatory responses.² Therefore, it is important to identify the presence of bacteria in DFUs. The gold standard for identifying the presence of bacteria in a chronic wound is tissue biopsy culture.³ However, because of its technical difficulty and invasiveness, swab cultures have been used as alternatives in several clinical trials. Swab culture is a relatively simple and less invasive method; however, its diagnostic validity in chronic wounds is unreliable.⁴ Moreover, both methods are time- and resource-consuming for bacterial culture and laboratory processes. 

To address these limitations, a bacterial fluorescence imaging device (MolecuLight i:X, MolecuLight Inc) was developed. This device utilizes the basic principle that when illuminated by violet light (405 nm), some bacteria containing endogenous porphyrins emit a red fluorescence signal (Figure 1A), while others such as Pseudomonas aeruginosa, which produce pyoverdine, emit a cyan fluorescence signal⁵ (Figure 1B). With a dim background light, this handheld device can capture real-time colored fluorescence imaging and has the advantages of being easy to learn, noninvasive, and able to provide immediate information about the presence of bacteria. 

Several studies have reported the diagnostic accuracy of the bacterial fluorescence imaging device^6-16; however, some limitations exist. In most studies, specimens are collected using the swab technique. Furthermore, most studies determined the diagnostic accuracy using comparison with quantitative polymerase chain reaction, which is not commonly used in the practical clinical setting, rather than using tissue culture methods. To the best of the current authors’ knowledge, no studies have been conducted on diabetic wounds alone. 

The purpose of this study was to investigate the diagnostic accuracy of a fluorescence imaging device for diabetic wounds using a tissue culture system which is the gold standard for identifying the presence of bacteria. To determine this, the sensitivity, specificity, PPV, and NPV of the device were measured, and possible differences in diagnostic accuracy between P aeruginosa and non-Pseudomonas bacteria were explored. Further, the results of previous studies were investigated.

Patients

This was a prospective, observational, single-center study. Patients who visited or were admitted to the diabetic wound center of Korea University Guro Hospital between April 2022 and July 2022 were included in this study. All patients had been previously diagnosed with diabetes and had persistent wounds for more than 2 weeks. A total of 35 consecutive patients (with 48 total wounds) were included in the study. In 8 patients, serial imaging and wound sampling were performed during admission. 

Research ethics and patient consent

This study was approved by the Institutional Review Board of the hospital, and consent from all patients was received.

Fluorescence imaging and tissue sampling

All wounds were cleaned and photographed to obtain fluorescence and normal images using the bacterial fluorescence imaging device. The images were compared to precisely identify the infectious area. During fluorescence imaging, wounds were entirely covered by DarkDrape (MolecuLight Inc), an imaging device accessory that provides proper ambient lighting conditions and precisely visualizes bacterial presence. Tissue biopsies were then performed aseptically. Tissue was collected from the colored regions of the fluorescence-positive wounds. For fluorescence-negative wounds, tissue was collected from the regions where the bacterial load was expected to be the highest visually. The tissues were cultured in thioglycollate broth for 48 hours at 37 °C, and bacterial identification and sensitivity tests were conducted according to the Clinical and Laboratory Standards Institute guidelines.¹⁷ All images were sent to the manufacturer of the imaging device, confirmed for image quality, and interpreted for bacterial presence.

Since non-Pseudomonas bacteria is shown with red color, and P aeruginosa is shown with cyan color, the culture outcomes were classified into non-Pseudomonas bacterial, Pseudomonas bacterial, both bacterial, and no-growth groups. Image outcomes were classified into red, cyan, both colors, and negative groups. 

Statistical analysis

By comparing the culture and image results, all wounds were evaluated as true positives (culture and image positive), true negatives (culture and image negative), false positives (culture negative but image positive), or false negatives (culture positive but image negative). Overall sensitivity, specificity, PPV, NPV, and accuracy were calculated using evaluated estimates. In addition, the McNemar test was conducted to find the strain among Pseudomonas or non-Pseudomonas bacteria that demonstrated higher diagnostic accuracy. Moreover, the sensitivity, specificity, PPV, and NPV obtained in this study were compared with those of previous studies using the chi-square test. Statistical analyses were performed using SPSS Statistics (version 23.0; IBM).

Patient characteristics

Of the 35 patients included in the study, 23 (66%) were male and 12 (34%) were female. The mean (± standard deviation) age was 60.2 ± 14.7 years. The average wound duration was 5.2 ± 5.6 months. The mean duration of diabetes was 16.9 ± 11.7 years. The most common underlying diseases included hypertension (49%, n = 17), diabetic peripheral neuropathy (40%, n = 14), and chronic kidney disease without hemodilaysis (20%, n = 7). Three patients underwent hemodialysis. Baseline patient information is summarized in Table 1. 

Image and culture outcomes

Of the tissue cultures, 81% (n = 39) were positive for bacterial growth and 19% (n = 9) were negative. Twenty-two bacterial species were found in culture-positive wounds. Staphylococcus aureus and P aeruginosa were the most commonly identified species (n = 9) (Table 2). Among the culture-positive results, non-Pseudomonas bacteria accounted for 77% (n = 30) of cases, followed by Pseudomonas bacteria (18%, n = 7), and both bacterial types (5%, n = 2) (Table 3).

For bacterial presence detected using fluorescence imaging, 60% (n = 29) of wounds were image-positive and 40% (n = 19) were image-negative. Among the image-positive wounds, 62% (n = 18) showed red, 31% (n = 9) showed cyan, and 7% (n = 2) showed both colors (Table 3).

Diagnostic accuracy

For identifying the presence of bacteria, the imaging device had a sensitivity, specificity, PPV, and NPV of 64.1%, 55.6%, 86.2%, and 26.3%, respectively; the results indicated an accuracy of 62.5%. For Pseudomonas identification, the device had a sensitivity, specificity, PPV, and NPV of 66.7%, 87.2%, 54.6%, and 91.9%, respectively, with an accuracy of 83.3%. For non-Pseudomonas bacterial identification, sensitivity, specificity, PPV, and NPV were 43.8%, 62.5%, 70.0%, and 35.7%, respectively, with an accuracy of 50.0% (Table 4). According to the McNemar test, the device exhibited higher diagnostic accuracy for Pseudomonas (P < .001). The chi-square test revealed that the sensitivity, specificity, PPV, and NPV were lower than those of previous studies. The specificity and NPV showed statistical significance (Table 5).

Identification of the bacterial burden in DFUs is key for appropriate treatment. An untreated bacterial burden can cause wound infections and may delay or prevent wound healing. According to the International Working Group on the Diabetic Foot and the Infectious Diseases Society of America, diabetic foot infection is defined as the presence of at least 2 inflammatory manifestations (erythema, pain, tenderness, or warmth). However, in patients with diabetes, the clinical signs and symptoms of foot infections may be significantly reduced. The presence of immunopathy (impaired inflammatory cell function) was higher in patients with diabetes than in those without diabetes. Clinical diagnosis of infection in patients with diabetes may be difficult. Furthermore, the poor immune-leukocyte response of individuals with diabetes may reduce the local inflammatory response and signs or symptoms of local infection. Moreover, systemic signs of toxicity, such as general malaise, numbness, nausea, anorexia, fever, and leukocytosis, may be absent or appear later, even in severe cases. Therefore, although diabetic wounds may lack clinical signs of infection, they may contain a high bacterial burden that results in subclinical infection or critical colonization, which may impair wound healing. Thus, the ability to diagnose infections in patients with diabetes may depend on the tissue culture method, but this issue remains debatable. Furthermore, tissue biopsy cultures require proficiency in terms of specimen sampling, and more than 48 hours are required for bacterial growth. Therefore, the fluorescence imaging device was developed for easy and rapid identification of bacterial burden. The bacterial fluorescence imaging device enables real-time verification of the presence of bacteria in a wound without the need for tools or preparation other than appropriate lighting conditions. This study aimed to determine the accuracy of the bacterial fluorescence imaging device compared to that of tissue biopsy culture results.

In this study, bacteria were cultured in tissue biopsied from 81% of patients, with S aureus and P aeruginosa being the most common strains identified. These results were consistent with those of a previous study on bacteria cultured from DFUs.¹⁸ The current study showed that the imaging device had a sensitivity, specificity, PPV, and NPV of 64.1%, 55.6%, 86.2%, and 26.3%, respectively. Previous studies on the average diagnostic accuracy of bacterial fluorescence imaging device showed a sensitivity of 74% (range, 59%–100%), specificity of 88% (range, 72%–100%), PPV of 91% (range, 67%–100%), and NPV of 53% (range, 17%–100%).^6-16 In addition, the current authors analyzed the results by dividing them into 2 groups: Pseudomonas and non-Pseudomonas. For non-Pseudomonas bacteria, the device had a sensitivity, specificity, PPV, and NPV of 43.8%, 62.5%, 70.0%, and 35.7%, respectively. For Pseudomonas, the device had a sensitivity, specificity, PPV, and NPV of 66.7%, 87.2%, 54.6%, and 91.9%, respectively. Overall, the detection of Pseudomonas was more accurate, similar to what was reported in previous studies.^6,9 

The current results showed a lower accuracy than that seen in previous studies. These discrepancies could be attributed to several factors. First, in most previous studies, samples were collected using swabs, whereas in the current study samples were collected using a tissue biopsy method. It is difficult to accurately identify pathogens when samples are collected using swabs because normal bacterial colonies may also be present in DFUs. Therefore, collecting specimens using a tissue biopsy method for chronic wounds, such as DFUs, has been recommended.⁴ In addition, since samples can only be collected from exposed wounds, pathogens present under the skin or necrotizing tissue cannot be detected using only swab techniques. Two studies that involved collecting samples from several types of wounds using the tissue biopsy method showed relatively lower sensitivities (59% and 73%) and NPVs (32% and 17%) than those reported in other studies, emphasizing the importance of this consideration.^6,10 Second, there may be errors in image interpretation. For fluorescence imaging, bacteria containing endogenous porphyrins and pyoverdine are excited by violet light and emit red and cyan signals, respectively. However, these signals are often not clearly visible because of the presence of skin, clots, and disinfectants or dyes around the wound. In addition, normal structures, including collagen and heme, may exhibit color signals like those of bacteria.¹⁹ To overcome this possibility, a normal image of the same view was captured along with the fluorescent image, and the 2 images were compared to increase the accuracy of the analysis. In addition, image interpretation was requested from the device manufacturer to further improve accuracy.

The advantages of the bacterial fluorescence imaging device are that it is noninvasive, does not require special proficiency, and can detect the presence or absence of bacteria in real time. However, the bacterial fluorescence imaging device has some disadvantages. First, only wounds with high bacterial density can be distinguished. The device detects bacteria at a density of greater than or equal to 10⁴ CFU/g.²⁰ The greater the bacterial burden in wounds, the higher the probability of developing infection and delayed wound healing. However, the cutoff value is debatable, as infection may occur even at a low bacterial density that the device cannot detect.^21,22 In cases of beta-hemolytic streptococci, an infection may even occur with a much lower bacteriological load than that required for other species.²³ Thus, this may influence and reduce the diagnostic accuracy. DFUs are more vulnerable to infection due to poor host resistance and tissue circulation, which can also affect the ability to diagnose infection using the bacterial fluorescence imaging device. There is also a limitation to the depth of the wound that can be captured. Because violet light may penetrate to a depth of 1.5 mm,²⁴ it can be difficult to accurately inspect wounds with a greater depth. DFUs may be accompanied by osteomyelitis; therefore, it is necessary to know whether the bacterial burden exists in deep tissues or bones.²⁵ Additionally, if the wound is not open, the violet light may not penetrate the deep tissues, which may reduce the diagnostic accuracy.

Considering these advantages and disadvantages, in actual clinical situations of treating diabetic feet, the bacterial fluorescence imaging device should be used as a screening tool to detect infection in asymptomatic DFUs, rather than as a replacement for tissue biopsy cultures. The device enables real-time detection of bacterial load and therefore can help with sampling and debridement in cases where the bacterial load is high. Further studies are required to better understand these clinical considerations.

This study has some limitations. First, it was conducted on a small number of cases (48 wounds). In this study, the difference in diagnostic accuracy of the device was confirmed for Pseudomonas and non-Pseudomonas. However, it could not be confirmed whether the diagnostic accuracy was different among other specific strains due to the insufficient number of cases. Second, this study showed low detection rate for polyinfections and anaerobic bacteria. DFUs are known to have polyinfection including anaerobic bacteria, but in this study, only 1 type of bacteria was identified in most wounds. This is because in the case of bacterial identification tests used in actual clinical practice, it is not easy to detect various bacteria including anaerobes due to cost and technical problems. To overcome these limitations, further research involving more cases and various strains will be needed.

The current authors investigated the diagnostic accuracy of a fluorescence imaging device in patients with DFUs using tissue culture techniques. From this study, the results showed the fluorescence imaging device helped detect bacterial bioburden; however, its accuracy may be lower than that reported in previous studies for patients with diabetes.