MisMatch: Calibrated Segmentation via Consistency on Differential Morphological Feature Perturbations With Limited Labels (2021-10-23T00:00:00.000000Z)

TL;DR

This paper proposes MisMatch, a semi-supervised segmentation framework based on the consistency between paired predictions which are derived from two differently learnt morphological feature perturbations and develops a 2D U-net based MisMatch framework and performs extensive cross-validation on a CT-based pulmonary vessel segmentation task.

Abstract

Semi-supervised learning (SSL) is a promising machine learning paradigm to address the ubiquitous issue of label scarcity in medical imaging. The state-of-the-art SSL methods in image classification utilise consistency regularisation to learn unlabelled predictions which are invariant to input level perturbations. However, image level perturbations violate the cluster assumption in the setting of segmentation. Moreover, existing image level perturbations are hand-crafted which could be sub-optimal. In this paper, we propose MisMatch, a semi-supervised segmentation framework based on the consistency between paired predictions which are derived from two differently learnt morphological feature perturbations. MisMatch consists of an encoder and two decoders. One decoder learns positive attention for foreground on unlabelled data thereby generating dilated features of foreground. The other decoder learns negative attention for foreground on the same unlabelled data thereby generating eroded features of foreground. We normalise the paired predictions of the decoders, along the batch dimension. A consistency regularisation is then applied between the normalised paired predictions of the decoders. We evaluate MisMatch on four different tasks. Firstly, we develop a 2D U-net based MisMatch framework and perform extensive cross-validation on a CT-based pulmonary vessel segmentation task and show that MisMatch statistically outperforms state-of-the-art semi-supervised methods. Secondly, we show that 2D MisMatch outperforms state-of-the-art methods on an MRI-based brain tumour segmentation task. We then further confirm that 3D V-net based MisMatch outperforms its 3D counterpart based on consistency regularisation with input level perturbations, on two different tasks including, left atrium segmentation from 3D CT images and whole brain tumour segmentation from 3D MRI images. Lastly, we find that the performance improvement of MisMatch over the baseline might originate from its better calibration. This also implies that our proposed AI system makes safer decisions than the previous methods.

References47 items

Learning Morphological Feature Perturbations for Calibrated Semi-Supervised Segmentation

Semi-Supervised Semantic Segmentation of Vessel Images using Leaking Perturbations

Inconsistency-Aware Uncertainty Estimation for Semi-Supervised Medical Image Segmentation

The Medical Segmentation Decathlon

A global benchmark of algorithms for segmenting the left atrium from late gadolinium-enhanced cardiac magnetic resonance imaging

DMNet: Difference Minimization Network for Semi-supervised Segmentation in Medical Images

Local and Global Structure-Aware Entropy Regularized Mean Teacher Model for 3D Left Atrium Segmentation

Semi-supervised Medical Image Segmentation through Dual-task Consistency

Guided Collaborative Training for Pixel-wise Semi-Supervised Learning

Learning To Pay Attention To Mistakes

Dual-Teacher: Integrating Intra-domain and Inter-domain Teachers for Annotation-efficient Cardiac Segmentation

Realistic Adversarial Data Augmentation for MR Image Segmentation

Semi-Supervised Learning for Fetal Brain MRI Quality Assessment with ROI consistency

A Global Benchmark of Algorithms for Segmenting Late Gadolinium-Enhanced Cardiac Magnetic Resonance Imaging

Semi-Supervised Semantic Segmentation With Cross-Consistency Training

FixMatch: Simplifying Semi-Supervised Learning with Consistency and Confidence

PyTorch: An Imperative Style, High-Performance Deep Learning Library

ReMixMatch: Semi-Supervised Learning with Distribution Alignment and Augmentation Anchoring

Randaugment: Practical automated data augmentation with a reduced search space

Multi-Task Attention-Based Semi-Supervised Learning for Medical Image Segmentation

Uncertainty-aware Self-ensembling Model for Semi-supervised 3D Left Atrium Segmentation

Semi-supervised semantic segmentation needs strong, varied perturbations

CutMix: Regularization Strategy to Train Strong Classifiers With Localizable Features

Curriculum semi-supervised segmentation

Label Propagation for Deep Semi-Supervised Learning

Semi-Supervised Brain Lesion Segmentation with an Adapted Mean Teacher Model

Semi-Supervised and Task-Driven Data Augmentation

Semi-supervised Skin Lesion Segmentation via Transformation Consistent Self-ensembling Model

There Are Many Consistent Explanations of Unlabeled Data: Why You Should Average

Revisiting Dilated Convolution: A Simple Approach for Weakly- and Semi-Supervised Semantic Segmentation

On Calibration of Modern Neural Networks

Mean teachers are better role models: Weight-averaged consistency targets improve semi-supervised deep learning results

Understanding the Effective Receptive Field in Deep Convolutional Neural Networks

V-Net: Fully Convolutional Neural Networks for Volumetric Medical Image Segmentation

Residual Networks Behave Like Ensembles of Relatively Shallow Networks

Automatic Pulmonary Artery-Vein Separation and Classification in Computed Tomography Using Tree Partitioning and Peripheral Vessel Matching

The Multimodal Brain Tumor Image Segmentation Benchmark (BRATS)

U-Net: Convolutional Networks for Biomedical Image Segmentation

Adam: A Method for Stochastic Optimization

Semi-supervised Learning with Deep Generative Models

Semi-supervised Learning by Entropy Minimization

The Comparison and Evaluation of Forecasters.

TO APPEAR tional

Comput Vis Pattern Recognit (CVPR)

Realistic Evaluation of Semi-Supervised Learning Algorithms

Pseudo-Label : The Simple and Efficient Semi-Supervised Learning Method for Deep Neural Networks

applicable license agreement with IEEE. Restrictions apply