OBJECTIVE  To construct and validate an image super-resolution reconstruction model suitable for indocyanine green （ICG） fluorescence imaging-guided debridement of postoperative infections. This technology enhances the image quality and resolution of fluorescence imaging, thereby improving the accuracy of postoperative infection debridement.

METHODS  The imaging data from 79 patients who underwent surgical site infection debridement （SSI） debridement guided by ICG at the Affiliated Hospital of Southwest Medical University from Jan. 2023 to Dec. 2024 were studied. Images were recorded at five time points （2, 4, 6, 8 and 10 hours） after ICG injection, and the data were divided in a ratio of 47:32. Enhanced deep super-resolution neural network public datasets were employed to train 2× and 4× super-resolution models, which were subsequently adapted for ICG fluorescence imaging. Quantitative metrics of the super-resolution ICG fluorescence images were calculated to evaluate the improvement in image quality. The accuracy of super-resolution images for debridement was assessed through annotation consistency and subjective image quality scoring.

RESULTS  Both 2× and 4× super-resolution reconstructions significantly improved peak signal-to-noise ratio, structural similarity and boundary gradient intensity in ICG fluorescence images. Annotation consistency confirmed the method's effectiveness in enhancing the discernibility of infection boundaries, while subjective quality scores of images validated its utility in debridement decision-making.

CONCLUSIONS  ICG fluorescence imaging for SSI debridement often suffers from blurred tissue boundaries and low image quality. Super-resolution reconstruction technique provides clearer tissue boundary images, thereby enhancing the accuracy of SSI debridement. Optimal image quality for debridement is achieved 6–8 hours after ICG injection.