InformAI's manuscript entitled nnU-Net Deep Learning Method for Segmenting Parenchyma and Determining Liver Volume from Computed Tomography Images has been published in the Annals of Surgery Open journal. This work highlights the adaptation of the nnU-Net to solve the complex problem of organ segmentation to facilitate appropriate size matching prior to surgical organ transplantation.
nnU-net is an extension of the highly impactful U-Net deep learning architecture which has been validated as an accurate automatic 3D image segmentation tool. nnU-Net builds upon the previous model, adding impressive object identification, enhanced image segmentation capabilities, and better generalizability.
Solid-organ transplantations are surgically-complex procedures. To achieve successful surgical outcomes, transplantation teams must appropriately consider numerous features for both an organ donor and a patient recipient. Of these considerations, the size and shape of an offered organ can be relevant. Size mismatches in which a donated organ is relatively large or a recipient has anatomical cavity limitations may prevent a transplantation procedure from occurring, or may lead to an adverse clinical outcome. In this way, estimation of donor organ volume and recipient abdominal cavity is essential. Deep learning methods allow optimal segmentation of organs from computed tomography (CT) scans, offering higher accuracy and precision over currently utilized methods.