Biomedical Image Registration, Domain Generalisation and Out-of-Distribution Analysis: MICCAI 2021 Challenges: MIDOG 2021, MOOD 2021, and Learn2Reg 2021, Held in Conjunction with MICCAI 2021, Strasbourg, France, September 27–October 1, 2021, ProceedingsMarc Aubreville, David Zimmerer, Mattias Heinrich This book constitutes three challenges that were held in conjunction with the 24th International Conference on Medical Image Computing and Computer-Assisted Intervention, MICCAI 2021, which was planned to take place in Strasbourg, France but changed to an online event due to the COVID-19 pandemic. The peer-reviewed 18 long and 9 short papers included in this volume stem from the following three biomedical image analysis challenges:
The challenges share the need for developing and fairly evaluating algorithms that increase accuracy, reproducibility and efficiency of automated image analysis in clinically relevant applications. |
Contents
| 5 | |
| 14 | |
DomainRobust Mitotic Figure Detection with Style Transfer | 23 |
TwoStep Domain Adaptation for Mitotic Cell Detection in Histopathology Images | 32 |
DomainSpecific CycleGAN Augmentation Improves Domain Generalizability for Mitosis Detection | 40 |
StainRobust Mitotic Figure Detection for the Mitosis Domain Generalization Challenge | 48 |
Simple and Robust Mitosis Detection | 53 |
Multisource Domain Adaptation Using Gradient Reversal Layer for Mitotic Cell Detection | 58 |
Selfsupervised Medical OutofDistribution Using UNet Vision Transformers | 104 |
Unsupervised OutofDistribution Detection and Localization for Medical Volumes | 111 |
Detecting Outliers by Learning to Learn from Selfsupervision | 119 |
AutoSeg Steering the Inductive Biases for Automatic Pathology Segmentation | 127 |
L2R | 136 |
Deformable Registration of Brain MR Images via a Hybrid Loss | 141 |
Fraunhofer MEVIS Image Registration Solutions for the Learn2Reg 2021 Challenge | 147 |
Unsupervised Volumetric Displacement Fields Using Cost Function Unrolling | 153 |
A Strategy for the MItosis DOmain Generalization MIDOG Challenge | 62 |
Detecting Mitosis Against Domain Shift Using a Fused Detector and Deep Ensemble Classification Model for MIDOG Challenge | 68 |
Domain Adaptive Cascade RCNN for MItosis DOmain Generalization MIDOG Challenge | 73 |
Domain Generalisation for Mitosis Detection Exploting Preprocessing Homogenizers | 77 |
Cascade RCNN for MIDOG Challenge | 81 |
SkUnet Model with Fourier Domain for Mitosis Detection | 86 |
MOOD | 91 |
Selfsupervised 3D OutofDistribution Detection via Pseudoanomaly Generation | 95 |
Conditional Deep Laplacian Pyramid Image Registration Network in Learn2Reg Challenge | 161 |
The Learn2Reg 2021 MICCAI Grand Challenge PIMed Team | 168 |
Fast 3D Registration with Accurate Optimisation and Little Learning for Learn2Reg 2021 | 174 |
Progressive and CoarsetoFine Network for Medical Image Registration Across Phases Modalities and Patients | 180 |
Semisupervised Multilevel Symmetric Image Registration Method for Magnetic Resonance Whole Brain Images | 186 |
| 192 | |
Other editions - View all
Common terms and phrases
abdominal algorithm annotations anomaly detection approach arXiv preprint Aubreville autoencoder baseline bounding box brain breast cancer Cham Check for updates classification color Computer Vision consists cost function CT scans CycleGAN data augmentation dataset deep learning deformation field detector Dice different scanners domain adversarial domain generalization challenge domain shift evaluation F1 score few-shot final test set gradients Hamamatsu Hamamatsu XR histopathology images hyperparameter IEEE Image Anal image registration improve input image labels Learn2Reg challenge learning rate Leica GT450 LNCS loss lung Machine Learning meta learning MICCAI MIDOG challenge mitosis detection MItosis DOmain module moving image neural networks optical flow optimization overfitting patches pathology pixels prediction preliminary test set random RetinaNet robust samples scanners self-supervised learning slices slide Springer Nature Switzerland style transfer Switzerland AG 2022 target task tion training data training images training set U-Net unseen unsupervised Unsupervised learning validation set voxel