Abstract:Accurate segmentation of subcortical structures is an important task in quantitative brain image analysis. Convolutional neural networks (CNNs) have achieved remarkable results in medical image segmentation. However, due to the difficulty of acquiring high-quality annotations of brain subcortical structures, learning segmentation networks using noisy annotations is an inevitable topic. A common practice is to select images or pixels with reliable annotations for training, which usually may not make full use of the information from the training samples, thus affecting the performance of the learned segmentation model. To address the above problem, in this work, we propose a novel robust learning method and denote it as uncertainty-reliability awareness learning (URAL), which can make sufficient use of all training pixels. At each training iteration, the proposed method first selects training pixels with reliable annotations from the set of pixels with uncertain network prediction, by utilizing a small clean validation set following a meta-learning paradigm. Meanwhile, we propose the online prototypical soft label correction (PSLC) method to estimate the pseudo-labels of label-unreliable pixels. Then, the segmentation loss of label-reliable pixels and the semi-supervised segmentation loss of label-unreliable pixels are used to calibrate the total segmentation loss. Finally, we propose a category-wise contrastive regularization to learn compact feature representations of all uncertain training pixels. Comprehensive experiments are performed on two publicly available brain MRI datasets. The proposed method achieves the best Dice scores and MHD values on both datasets compared to several recent state-of-the-art methods under all label noise settings. Our code is available at https://github.com/neulxlx/URAL.

Constrained unsupervised anomaly segmentation

An Accurate Unsupervised Liver Lesion Detection Method Using Pseudo-lesions

Unsupervised Anomaly Segmentation for Brain Lesions Using Dual Semantic-Manifold Reconstruction.

ASC-Net : Adversarial-based Selective Network for Unsupervised Anomaly Segmentation

Anomaly-Aware Semantic Segmentation by Leveraging Synthetic-Unknown Data

ASC-Net: Unsupervised Medical Anomaly Segmentation Using an Adversarial-based Selective Cutting Network

Pull & Push: Leveraging Differential Knowledge Distillation for Efficient Unsupervised Anomaly Detection and Localization

Unsupervised Anomaly Segmentation Via Multilevel Image Reconstruction and Adaptive Attention-Level Transition

Unsupervised Anomaly Segmentation using Image-Semantic Cycle Translation

AnoSeg: Anomaly Segmentation Network Using Self-Supervised Learning

IterMask2: Iterative Unsupervised Anomaly Segmentation via Spatial and Frequency Masking for Brain Lesions in MRI

Unsupervised Lesion Detection via Image Restoration with a Normative Prior

Segment Any Anomaly without Training via Hybrid Prompt Regularization

Learning to segment subcortical structures from noisy annotations with a novel uncertainty-reliability aware learning framework

Synomaly Noise and Multi-Stage Diffusion: A Novel Approach for Unsupervised Anomaly Detection in Ultrasound Imaging

Weakly Supervised Brain Lesion Segmentation via Attentional Representation Learning

Point-supervised Segmentation of Microscopy Images and Volumes via Objectness Regularization

Attention-Guided Perturbation for Unsupervised Image Anomaly Detection

Anatomically-aware uncertainty for semi-supervised image segmentation

Spatial-contextual variational autoencoder with attention correction for anomaly detection in retinal OCT images

Normal Image Guided Segmentation Framework for Unsupervised Anomaly Detection