Detection of flap malperfusion after microsurgical tissue reconstruction using hyperspectral imaging and machine learning.

Hyperspectral imaging (HSI) has shown significant diagnostic potential for both intra- and postoperative perfusion assessment. The purpose of this study was to combine machine learning and neural networks with HSI to develop a method for detecting flap malperfusion after microsurgical tissue reconstruction. Data records were analysed to assess the occurrence of flap loss after microsurgical procedures. A total of 59 free flaps were recorded, ten of which demonstrated postoperative malperfusion, leading to necrosis. Several supervised classification algorithms were evaluated to differentiate impaired perfusion from healthy tissue via HSI recordings. The best flap classification performance was observed using a convolutional neural network using HSI based perfusion parameters within 72 h after surgery, with an area under the curve of 0.82 ± 0.05, a sensitivity of 70% ± 33%, a specificity of 76% ± 26%, and an F1 score of 68% ± 28%. HSI combined with artificial intelligence approaches in diagnostic tools could significantly improve the detection of postoperative malperfusion and potentially increase flap salvage rates.