SRS279 - Using nnU-Net to segment MRI images of the foot in patients with diabetic foot disease

Abstract Introduction Diabetic foot disease (DFD) is a complex disease and is associated with lower limb amputation. Magnetic resonance imaging (MRI) is commonly used in patients with DFD. Early detection of surrogate markers on MRI may help to predict the risk of limb loss. No-new-U-Net (nnU-Net) is a convolutional neural network designed for semantic segmentation in medical imaging. Aims To explore the use of deep learning models to identify predictors on MRI for limb loss in patients with DFD. Methods 824 three-dimensional images of MRI scans of the foot were manually segmented at the base of the 1st metatarsal. nnU-Net was applied to segment using the same ratio. The pseudo fat fraction was also calculated by measuring the muscle and plantar fat signal intensities. Results The mean DICE score, which is a ratio of the ground truth and the predicted segmentation was 0.4 (on a scale of zero to one) and therefore was not an accurate model. The range of DICE scores varied from 0.0 to 0.88 which highlights the potential for its use in future. Conclusions The use of nnU-Net in MRI images of the foot in patients with DFD requires further exploration as it has potential.