Development and validation of an algorithm to predict the treatment modality of burn wounds using thermographic scans: Prospective cohort study

Abstract

Background: The clinical evaluation of a burn wound alone may not be adequate to predict the severity of the injury nor to guide clinical decision making. Infrared thermography provides information about soft tissue viability and has previously been used to assess burn depth. The objective of this study was to determine if temperature differences in burns assessed by infrared thermography could be used predict the treatment modality of either healing by re-epithelization, requiring skin grafts, or requiring amputations, and to validate the clinical predication algorithm in an independent cohort.

Methods and findings: Temperature difference (ΔT) between injured and healthy skin were recorded within the first three days after injury in previously healthy burn patients. After discharge, the treatment modality was categorized as re-epithelization, skin graft or amputation. Potential confounding factors were assessed through multiple linear regression models, and a prediction algorithm based on the ΔT was developed using a predictive model using a recursive partitioning Random Forest machine learning algorithm. Finally, the prediction accuracy of the algorithm was compared in the development cohort and an independent validation cohort. Significant differences were found in the ΔT between treatment modality groups. The developed algorithm correctly predicts into which treatment category the patient will fall with 85.35% accuracy. Agreement between predicted and actual treatment for both cohorts was weighted kappa 90%.

Conclusion: Infrared thermograms obtained at first contact with a wounded patient can be used to accurately predict the definitive treatment modality for burn patients. This method can be used to rationalize treatment and streamline early wound closure.

Conflict of interest statement

The authors have declared that no competing interests exist.

Figures

Fig 1. Study flowchart.

Fig 2. Thermograms and treatment groups.

Clinical images (A to C) and thermograms (A1 to C1) were obtained during the first three days after the injury, the patients were followed up until discharged, and their outcome classified as healed by conservative treatment, skin graft or amputation. Significant differences in the temperature difference (TD) between injured and healthy tissue across all groups were found (D). Panels A and A1 represent an injury that healed by epithelization, B and B1 one that required a skin graft (note the colder region on the ulnar surface of the forearm), and C and C1 an injury that required amputation. Noteworthy in this last image, even though all the leg skin looks charred in the clinical image, the thermogram suggested that only the feet were non-viable.

Fig 3. Clinical decision algorithm based on thermograms.

Through Random Forest algorithms, the following clinical decision rule was developed: if a patient presents a temperature difference (ΔT) of >5.0°C, he will require amputation of the affected limb. If the ΔT is 3.0°C, he will require a skin graft; and if the ΔT is

Fig 4. Unsupervised clustering of datapoints.

Unsupervised k-means clustering of ΔT values of the development cohort was used to confirm the results of the Random Forest algorithm. In the x-axis of the graph, the three actual treatments used can be seen. Three clusters of datapoints arise based on the grouping of similar ΔT values. On the re-epithelization group, only one datapoint (green) lies outside of its cluster, while in the skin graft group, two datapoints (red) lie outside of their cluster. In the amputation group, all datapoints have been clustered together (blue). This technique supports the notion that a ΔT value of 3 and 5 correctly discriminates between treatment groups, regardless of the surgeon’s decision.

Fig 5. Use of thermography for clinical decision making.

Consultation to the burn clinic was requested for a 4-week old infant that had sustained a complete partial thickness burn to his left foot from a heat radiator (A). The paediatric surgeon on call had decided to hospitalize the patient and treat him with a skin graft based on the clinical characteristic of the wound but requested a second opinion to our clinic. The thermographic image showed a ΔT value of 1.8 (B), so conservative management with outpatient management and daily visits to the emergency department to monitor the wound was advised. After seven days of treatment, the wound showed signs of re-epithelization and adequate tissue perfusion. The patient evolution was satisfactory and was discharged from the burn clinic two weeks after the injury.

Fig 6. Use of thermography to guide amputation levels.

Thermographic imaging can be used as an adjunct to determine amputation levels in severely burned patients. A 24-year old patient with fourth degree burns in approximately 50% of his body surface area because of direct fire was admitted to the burn unit (A). Both legs were severally burned and charred to the clinical inspection. The thermographic image showed progressing ΔT values from 12.7 distally to 1.2 proximally (B, C). A ΔT = 3 was used as a guide to select the amputation level, as it marks the limit for skin grafts and tissue viability. The amputation level is pointed by the forceps on panel A and B. Beyond this level, a sharp decrease in temperature values can be observed in panel C. The patient underwent supracondylar amputation of both legs, as well as tangential excision of all charred skin and was treated with skin grafts. At the moment of publication of this vignette, the patient is still being treated at the burn unit. This approach could also be potentially applied in patients with peripheral vascular disease to promote limb salvage or select optimal levels to create flaps for wound management and future prosthesis fitting.