Indocyanine Green Fluorescence Angiography in Plastic and Colorectal Surgery

Background: Indocyanine green fluorescence angiography (ICGFA) is gaining popularity for the assessment of reconstructive flap perfusion intraoperatively and is the most studied technology for colorectal resections. However, they are not widely available and used due to uncertainty in value and interpretation. Objectives: (1) to evaluate clinical and cost benefits of ICGFA in reconstructive procedures, (2) to
explore factors affecting widespread adoption and the level of benefit needed for routine use of ICGFA in both plastic and colorectal surgery, (3) and to develop and evaluate artificial intelligence methods that represent digitally how expert users would interpret ICGFA signalling for flap reconstructions and colorectal resections. Methods: I conducted a systematic review and meta-analysis to evaluate clinical and economic efficacy of ICGFA for plastic and reconstructive surgery. I then surveyed plastic and colorectal surgeons to identify their main challenges and level of benefit needed. Finally, I developed and tested artificial intelligence algorithms for expert ICGFA interpretations. Results: A meta-analysis of 25 studies found significant reductions in mastectomy skin flap necrosis, fat necrosis, infection rate, and re-operation in breast reconstruction, and partial flap loss in head and neck reconstructions. 7 studies reported cost savings with flap surgeries and breast reconstructions. Among 20 plastic surgeons, cost, training, added operating time, and other barriers were identified. 200 colorectal surgeons (across 36 countries) rated strong clinical evidence and standard protocol as important factors affecting routine adoption. The required benefits for colorectal surgery were savings of €250-500 per case and a number needed to treat between 20 and 40. AI-ICGFA for flap procedures was successful in predicting the need and extent of trimming here necessary. Maximum fluorescence intensity was identified as the most important predictive factor with a threshold of 22.1 grayscale units for excision decisions. AI-ICGFA for colorectal resections was also successful with sequence approach proving discriminative regarding expert ICGFA interpretation. It was generalisable to other imaging systems and expert surgeons from other institutions. The AI algorithm was also able to learn from multiple surgeons, and therefore, the predictions represent a shared view among expert ICGFA surgeons, reassuring the user. Conclusion: As evidence in ICGFA in plastic surgery grows, more attention is needed on identified factors. In colorectal surgery, the level of benefit expected seems in line with the current evidence. The AI-ICGFA expert interpretation representer holds great promise in solving the ICG problem and will now be key to its widespread adoption. Next steps include pre-market investigation of AI-software hardware device.