Computer Vision News - May 2023

30 Best Oral ISBI 2023 performance! ” The team’s focus on POD obliteration is the first step in their project. They plan to focus on other symptoms in the future, with the ultimate goal being that one day there will be a functioning system able to diagnose endometriosis for any sign based on any modality. Winning an award at a prestigious international event is an outstanding achievement for Yuan in more ways than one. Born in Southwest China (“ Where the panda comes from! ”), not only is she working in a non-native language, but she is also an oral presentation winner among Yuan used data augmentation techniques to create more samples from a small MRI dataset of only 89 cases. Preparing the data, she used the 3D AHE technique to enhance the definition of edges, while a pre-training method called 3D MAE helped the student encoder learn better about MRIs and achieve a promising reconstruction. “ The pre-training and fine-tuning datasets used in the model are very different, but we expect the pre-training to learn some patterns for normal or common MRIs, ” she explains. “ With this pre-training, we see a big improvement in our model’s B E OR I S 2 0 IEEE ISBI