Generalised Dice overlap as a deep learning loss function for highly unbalanced segmentations

Abstract

Deep-learning has proved in recent years to be a powerful tool for image analysis and is now widely used to segment both 2D and 3D medical images. Deep-learning segmentation frameworks rely not only on the choice of network architecture but also on the choice of loss function. When the segmentation process targets rare observations, a severe class imbalance is likely to occur between candidate labels, thus resulting in sub-optimal performance. In order to mitigate this issue, strategies such as the weighted cross-entropy function, the sensitivity function or the Dice loss function, have been proposed. In this work, we investigate the behavior of these loss functions and their sensitivity to learning rate tuning in the presence of different rates of label imbalance across 2D and 3D segmentation tasks. We also propose to use the class re-balancing properties of the Generalized Dice overlap, a known metric for segmentation assessment, as a robust and accurate deep-learning loss function for unbalanced tasks.

References in corpus (4)

• Efficient Multi-Scale 3D CNN with Fully Connected CRF for Accurate Brain Lesion Segmentation
• Brain Tumor Segmentation with Deep Neural Networks
• On the Compactness, Efficiency, and Representation of 3D Convolutional Networks: Brain Parcellation as a Pretext Task
• Deep Learning for Medical Image Segmentation

Cited by in corpus (235)

• ResUNet-a: a deep learning framework for semantic segmentation of remotely sensed data
• Recent advances and clinical applications of deep learning in medical image analysis
• A survey of loss functions for semantic segmentation
• Medical Image Segmentation Using Deep Learning: A Survey
• Automated brain extraction of multi-sequence MRI using artificial neural networks
• A review: Deep learning for medical image segmentation using multi-modality fusion
• NiftyNet: a deep-learning platform for medical imaging
• AnatomyNet: Deep Learning for Fast and Fully Automated Whole-volume Segmentation of Head and Neck Anatomy
• Automatic Brain Tumor Segmentation using Cascaded Anisotropic Convolutional Neural Networks
• Boundary loss for highly unbalanced segmentation
• Weakly-Supervised Convolutional Neural Networks for Multimodal Image Registration
• Modified U-Net (mU-Net) with Incorporation of Object-Dependent High Level Features for Improved Liver and Liver-Tumor Segmentation in CT Images
• Optimization for Medical Image Segmentation: Theory and Practice when evaluating with Dice Score or Jaccard Index
• Federated Learning Enables Big Data for Rare Cancer Boundary Detection
• Confidence Calibration and Predictive Uncertainty Estimation for Deep Medical Image Segmentation
• Association of genomic subtypes of lower-grade gliomas with shape features automatically extracted by a deep learning algorithm
• Optimizing the Dice Score and Jaccard Index for Medical Image Segmentation: Theory & Practice
• Deep Learning Approaches for Data Augmentation in Medical Imaging: A Review
• A Topological Loss Function for Deep-Learning based Image Segmentation using Persistent Homology
• Building extraction with vision transformer
• Asymmetric Loss Functions and Deep Densely Connected Networks for Highly Imbalanced Medical Image Segmentation: Application to Multiple Sclerosis Lesion Detection
• Dual Cross-Attention for Medical Image Segmentation
• 3D Segmentation with Exponential Logarithmic Loss for Highly Unbalanced Object Sizes
• Fully-automated Body Composition Analysis in Routine CT Imaging Using 3D Semantic Segmentation Convolutional Neural Networks
• An automatic multi-tissue human fetal brain segmentation benchmark using the Fetal Tissue Annotation Dataset
• Gaze-in-wild: A dataset for studying eye and head coordination in everyday activities
• Attention-Guided Version of 2D UNet for Automatic Brain Tumor Segmentation
• RITnet: Real-time Semantic Segmentation of the Eye for Gaze Tracking
• Two-Stage Mesh Deep Learning for Automated Tooth Segmentation and Landmark Localization on 3D Intraoral Scans
• EllSeg: An Ellipse Segmentation Framework for Robust Gaze Tracking
• Automatic Ischemic Stroke Lesion Segmentation from Computed Tomography Perfusion Images by Image Synthesis and Attention-Based Deep Neural Networks
• Relational Modeling for Robust and Efficient Pulmonary Lobe Segmentation in CT Scans
• A Deep Learning Framework for Spatiotemporal Ultrasound Localization Microscopy
• Morpheus: A Deep Learning Framework For Pixel-Level Analysis of Astronomical Image Data
• Generalised Wasserstein Dice Score for Imbalanced Multi-class Segmentation using Holistic Convolutional Networks
• UNet++: Redesigning Skip Connections to Exploit Multiscale Features in Image Segmentation
• Hybrid-Segmentor: A Hybrid Approach to Automated Fine-Grained Crack Segmentation in Civil Infrastructure
• Convolutional Neural Networks based Focal Loss for Class Imbalance Problem: A Case Study of Canine Red Blood Cells Morphology Classification
• Effective Training Strategies for Deep-learning-based Precipitation Nowcasting and Estimation
• Deep learning for brain metastasis detection and segmentation in longitudinal MRI data
• Generalized Wasserstein Dice Score, Distributionally Robust Deep Learning, and Ranger for brain tumor segmentation: BraTS 2020 challenge
• A volumetric deep Convolutional Neural Network for simulation of mock dark matter halo catalogues
• Automatic Segmentation of Vestibular Schwannoma from T2-Weighted MRI by Deep Spatial Attention with Hardness-Weighted Loss
• Advanced sleep spindle identification with neural networks
• PocketNet: A Smaller Neural Network for Medical Image Analysis
• GaNDLF: A Generally Nuanced Deep Learning Framework for Scalable End-to-End Clinical Workflows in Medical Imaging
• RLAS-BIABC: A Reinforcement Learning-Based Answer Selection Using the BERT Model Boosted by an Improved ABC Algorithm
• DoubleU-NetPlus: A Novel Attention and Context Guided Dual U-Net with Multi-Scale Residual Feature Fusion Network for Semantic Segmentation of Medical Images
• Exploring Feature Representation Learning for Semi-supervised Medical Image Segmentation
• Deep learning in radiology: an overview of the concepts and a survey of the state of the art
• IFSS-Net: Interactive Few-Shot Siamese Network for Faster Muscle Segmentation and Propagation in Volumetric Ultrasound
• Dice Loss for Data-imbalanced NLP Tasks
• On the influence of Dice loss function in multi-class organ segmentation of abdominal CT using 3D fully convolutional networks
• Super-Resolved Microbubble Localization in Single-Channel Ultrasound RF Signals Using Deep Learning
• Pan-tumor CAnine cuTaneous Cancer Histology (CATCH) dataset
• K-Lane: Lidar Lane Dataset and Benchmark for Urban Roads and Highways
• Deep Learning-Guided Surface Characterization for Autonomous Hydrogen Lithography
• A Survey and Analysis on Automated Glioma Brain Tumor Segmentation and Overall Patient Survival Prediction
• SiamixFormer: a fully-transformer Siamese network with temporal Fusion for accurate building detection and change detection in bi-temporal remote sensing images
• Constrained domain adaptation for Image segmentation
• Deep Convolutional Encoder-Decoders with Aggregated Multi-Resolution Skip Connections for Skin Lesion Segmentation
• Segmentation Loss Odyssey
• Automated Segmentation of Pulmonary Lobes using Coordination-Guided Deep Neural Networks
• A comparative study of 2D image segmentation algorithms for traumatic brain lesions using CT data from the ProTECTIII multicenter clinical trial
• CaBuAr: California Burned Areas dataset for delineation
• DARTS: DenseUnet-based Automatic Rapid Tool for brain Segmentation
• Automated Segmentation and Volume Measurement of Intracranial Carotid Artery Calcification on Non-Contrast CT
• Deep Small Bowel Segmentation with Cylindrical Topological Constraints
• InfiNet: Fully Convolutional Networks for Infant Brain MRI Segmentation
• SemiCurv: Semi-Supervised Curvilinear Structure Segmentation
• RRCANet: Recurrent Reusable-Convolution Attention Network for Infrared Small Target Detection
• Deep Learning for Segmentation of Cracks in High-Resolution Images of Steel Bridges
• Macroscale fracture surface segmentation via semi-supervised learning considering the structural similarity
• Distribution-aware Margin Calibration for Semantic Segmentation in Images
• blob loss: instance imbalance aware loss functions for semantic segmentation
• Cardiac Segmentation on CT Images through Shape-Aware Contour Attentions
• Transformers-based architectures for stroke segmentation: A review
• Investigating Shift-Variance of Convolutional Neural Networks in Ultrasound Image Segmentation
• Theoretical analysis and experimental validation of volume bias of soft Dice optimized segmentation maps in the context of inherent uncertainty
• Joint Learning of Blind Super-Resolution and Crack Segmentation for Realistic Degraded Images
• Deep morphological recognition of kidney stones using intra-operative endoscopic digital videos
• Deep, Deep Learning with BART
• Inversion of Time-Lapse Surface Gravity Data for Detection of 3D CO${}_2$ Plumes via Deep Learning
• Deep Multimodal Fusion of Data with Heterogeneous Dimensionality via Projective Networks
• Deep-XFCT: Deep learning 3D-mineral liberation analysis with micro X-ray fluorescence and computed tomography
• Acute ischemic stroke lesion segmentation in non-contrast CT images using 3D convolutional neural networks
• DefectNET: multi-class fault detection on highly-imbalanced datasets
• clDice -- A Novel Topology-Preserving Loss Function for Tubular Structure Segmentation
• Generalized Wasserstein Dice Loss, Test-time Augmentation, and Transformers for the BraTS 2021 challenge
• Optimization with soft Dice can lead to a volumetric bias
• Brain Tumor Segmentation using 3D-CNNs with Uncertainty Estimation
• Medical Instrument Segmentation in 3D US by Hybrid Constrained Semi-Supervised Learning
• Domain-Specific Deep Learning Feature Extractor for Diabetic Foot Ulcer Detection
• RIT-Eyes: Rendering of near-eye images for eye-tracking applications
• Be the Change You Want to See: Revisiting Remote Sensing Change Detection Practices
• EdgeSAM: Prompt-In-the-Loop Distillation for SAM
• Panoptic Perception: A Novel Task and Fine-grained Dataset for Universal Remote Sensing Image Interpretation
• Generative Model-Based Ischemic Stroke Lesion Segmentation
• Multi-organ segmentation: a progressive exploration of learning paradigms under scarce annotation
• Exploring The Limits Of Data Augmentation For Retinal Vessel Segmentation
• Iterative, Deep Synthetic Aperture Sonar Image Segmentation
• FU-net: Multi-class Image Segmentation Using Feedback Weighted U-net
• AdvSPADE: Realistic Unrestricted Attacks for Semantic Segmentation
• KAN You See It? KANs and Sentinel for Effective and Explainable Crop Field Segmentation
• Automated Olfactory Bulb Segmentation on High Resolutional T2-Weighted MRI
• AutoSourceID-Light. Fast Optical Source Localization via U-Net and Laplacian of Gaussian
• Multi-Task Generative Adversarial Network for Handling Imbalanced Clinical Data
• Terrain-Informed Self-Supervised Learning: Enhancing Building Footprint Extraction from LiDAR Data with Limited Annotations
• SegTHOR: Segmentation of Thoracic Organs at Risk in CT images
• ML-based Approaches for Wireless NLOS Localization: Input Representations and Uncertainty Estimation
• Superpixels and Graph Convolutional Neural Networks for Efficient Detection of Nutrient Deficiency Stress from Aerial Imagery
• Kidney segmentation using 3D U-Net localized with Expectation Maximization
• 3D-RETR: End-to-End Single and Multi-View 3D Reconstruction with Transformers
• Localization in the Crowd with Topological Constraints
• Multiclass Yeast Segmentation in Microstructured Environments with Deep Learning
• CT organ segmentation using GPU data augmentation, unsupervised labels and IOU loss
• On the Road with 16 Neurons: Mental Imagery with Bio-inspired Deep Neural Networks
• Proactive Prioritization of App Issues via Contrastive Learning
• Multi-class segmentation under severe class imbalance: A case study in roof damage assessment
• Spatial-temporal V-Net for automatic segmentation and quantification of right ventricles in gated myocardial perfusion SPECT images
• Topologically Faithful Multi-class Segmentation in Medical Images
• LapNet : Automatic Balanced Loss and Optimal Assignment for Real-Time Dense Object Detection
• SemSegLoss: A python package of loss functions for semantic segmentation
• Enhanced U-Net: A Feature Enhancement Network for Polyp Segmentation
• Lung infection and normal region segmentation from CT volumes of COVID-19 cases
• Deep Common Feature Mining for Efficient Video Semantic Segmentation
• U-Net with Hierarchical Bottleneck Attention for Landmark Detection in Fundus Images of the Degenerated Retina
• Collaborative Boundary-aware Context Encoding Networks for Error Map Prediction
• Improved HER2 Tumor Segmentation with Subtype Balancing using Deep Generative Networks
• Deep Residual 3D U-Net for Joint Segmentation and Texture Classification of Nodules in Lung
• A comparative study of semi- and self-supervised semantic segmentation of biomedical microscopy data
• Heuristical Comparison of Vision Transformers Against Convolutional Neural Networks for Semantic Segmentation on Remote Sensing Imagery
• LAMP: Large Deep Nets with Automated Model Parallelism for Image Segmentation
• Enhanced SCanNet with CBAM and Dice Loss for Semantic Change Detection
• The Dice loss in the context of missing or empty labels: Introducing $\Phi$ and $\epsilon$
• Breast mass segmentation based on ultrasonic entropy maps and attention gated U-Net
• A cascaded deep network for automated tumor detection and segmentation in clinical PET imaging of diffuse large B-cell lymphoma
• Cosmic-CoNN: A Cosmic Ray Detection Deep-Learning Framework, Dataset, and Toolkit
• Automatic airway segmentation from Computed Tomography using robust and efficient 3-D convolutional neural networks
• More unlabelled data or label more data? A study on semi-supervised laparoscopic image segmentation
• Few-Shot Connectivity-Aware Text Line Segmentation in Historical Documents
• DeepHIPS: A novel Deep Learning based Hippocampus Subfield Segmentation method
• Secondary Vertex Reconstruction with MaskFormers
• Improving Segmentation and Detection of Lesions in CT Scans Using Intensity Distribution Supervision
• Multitask Learning for Class-Imbalanced Discourse Classification
• Leishmaniasis Parasite Segmentation and Classification using Deep Learning
• Deep ensembles based on Stochastic Activation Selection for Polyp Segmentation
• SafeSteps: Learning Safer Footstep Planning Policies for Legged Robots via Model-Based Priors
• Globally Optimal Segmentation of Mutually Interacting Surfaces using Deep Learning
• Learn to Segment Retinal Lesions and Beyond
• Drawing the Line: Deep Segmentation for Extracting Art from Ancient Etruscan Mirrors
• UKAN-EP: Enhancing U-KAN with Efficient Attention and Pyramid Aggregation for 3D Multi-Modal MRI Brain Tumor Segmentation
• Segmentation of Head and Neck Organs at Risk Using CNN with Batch Dice Loss
• Partial annotations for the segmentation of large structures with low annotation cost
• Alleviating Class-wise Gradient Imbalance for Pulmonary Airway Segmentation
• Offset Curves Loss for Imbalanced Problem in Medical Segmentation
• Convolutional Recurrent Residual U-Net Embedded with Attention Mechanism and Focal Tversky Loss Function for Cancerous Nuclei Detection
• Learning Global and Local Features of Normal Brain Anatomy for Unsupervised Abnormality Detection
• OGNet: Salient Object Detection with Output-guided Attention Module
• Deepfake Image Generation for Improved Brain Tumor Segmentation
• Multi-task learning for joint weakly-supervised segmentation and aortic arch anomaly classification in fetal cardiac MRI
• Lung Segmentation in Chest X-rays with Res-CR-Net
• ICDAR 2021 Competition on Components Segmentation Task of Document Photos
• Improving Polyphonic Sound Event Detection on Multichannel Recordings with the Sørensen-Dice Coefficient Loss and Transfer Learning
• COVID-19 Infection Segmentation from Chest CT Images Based on Scale Uncertainty
• SCNet: A Generalized Attention-based Model for Crack Fault Segmentation
• Towards Tumour Graph Learning for Survival Prediction in Head & Neck Cancer Patients
• Comparison of Cross-Entropy, Dice, and Focal Loss for Sea Ice Type Segmentation
• Towards Investigating Residual Hearing Loss: Quantification of Fibrosis in a Novel Cochlear OCT Dataset
• CXR-Net: An Artificial Intelligence Pipeline for Quick Covid-19 Screening of Chest X-Rays
• FROG: A new people detection dataset for knee-high 2D range finders
• Ensemble Methods for Multi-Organ Segmentation in CT Series
• Decomposing Normal and Abnormal Features of Medical Images into Discrete Latent Codes for Content-Based Image Retrieval
• A Comparison of Deep Learning Convolution Neural Networks for Liver Segmentation in Radial Turbo Spin Echo Images
• AutoML Segmentation for 3D Medical Image Data: Contribution to the MSD Challenge 2018
• Segmentation task for fashion and apparel
• Learn distributed GAN with Temporary Discriminators
• Confidence-guided Lesion Mask-based Simultaneous Synthesis of Anatomic and Molecular MR Images in Patients with Post-treatment Malignant Gliomas
• Eye Semantic Segmentation with a Lightweight Model
• Flood Detection On Low Cost Orbital Hardware
• Deep Learning for Automated Medical Image Analysis
• Structure-aware scale-adaptive networks for cancer segmentation in whole-slide images
• Affordance segmentation of hand-occluded containers from exocentric images
• On-the-fly Prediction of Protein Hydration Densities and Free Energies using Deep Learning
• Disentangled Representations for Domain-generalized Cardiac Segmentation
• Learning Non-Unique Segmentation with Reward-Penalty Dice Loss
• Segmenting Ships in Satellite Imagery With Squeeze and Excitation U-Net
• Waveform-Specific Performance of Deep Learning-Based Super-Resolution for Ultrasound Contrast Imaging
• Morphologically-Aware Consensus Computation via Heuristics-based IterATive Optimization (MACCHIatO)
• Comparing Deep Learning strategies for paired but unregistered multimodal segmentation of the liver in T1 and T2-weighted MRI
• Few-Shot Segmentation of Historical Maps via Linear Probing of Vision Foundation Models
• Instance Segmentation XXL-CT Challenge of a Historic Airplane
• AttentionBoost: Learning What to Attend by Boosting Fully Convolutional Networks
• Out-of-distribution data supervision towards biomedical semantic segmentation
• Attentional Triple-Encoder Network in Spatiospectral Domains for Medical Image Segmentation
• A drone-based framework for coral habitat mapping via weakly supervised segmentation
• Lesion Mask-based Simultaneous Synthesis of Anatomic and MolecularMR Images using a GAN
• Detection and Prediction of Nutrient Deficiency Stress using Longitudinal Aerial Imagery
• Recurrent convolutional neural networks for mandible segmentation from computed tomography
• A First Look: Towards Explainable TextVQA Models via Visual and Textual Explanations
• Fusing Forces: Deep-Human-Guided Refinement of Segmentation Masks
• Adapting to Unseen Vendor Domains for MRI Lesion Segmentation
• Automated Mosaic Tesserae Segmentation via Deep Learning Techniques
• Enhancing Corpus Callosum Segmentation in Fetal MRI via Pathology-Informed Domain Randomization
• Validation of Various Normalization Methods for Brain Tumor Segmentation: Can Federated Learning Overcome This Heterogeneity?
• Real-time Segmentation and Facial Skin Tones Grading
• Voxel-level Siamese Representation Learning for Abdominal Multi-Organ Segmentation
• Efficient Scene Text Detection with Textual Attention Tower
• Development and evaluation of intraoperative ultrasound segmentation with negative image frames and multiple observer labels
• Seeing as Humans Do: Learning from Motion to Segment Anything Without Supervision
• Ensemble of Multi-sized FCNs to Improve White Matter Lesion Segmentation
• Technical report: Kidney tumor segmentation using a 2D U-Net followed by a statistical post-processing filter
• SFTrack++: A Fast Learnable Spectral Segmentation Approach for Space-Time Consistent Tracking
• Tissue characterization based on the analysis on i3DUS data for diagnosis support in neurosurgery
• 3D-OOCS: Learning Prostate Segmentation with Inductive Bias
• CSSegNet: Fine-Grained Cardiac Structures Segmentation Using Dilated Pyramid Pooling in U-net
• Spinal Cord Gray Matter-White Matter Segmentation on Magnetic Resonance AMIRA Images with MD-GRU
• PGD-UNet: A Position-Guided Deformable Network for Simultaneous Segmentation of Organs and Tumors
• Soft labeling by Distilling Anatomical knowledge for Improved MS Lesion Segmentation
• Deep Label Fusion: A 3D End-to-End Hybrid Multi-Atlas Segmentation and Deep Learning Pipeline
• Segmentation of Planning Target Volume in CT Series for Total Marrow Irradiation Using U-Net
• Soft Dice Confidence: A Near-Optimal Confidence Estimator for Selective Prediction in Semantic Segmentation
• High-level Prior-based Loss Functions for Medical Image Segmentation: A Survey
• Cell abundance aware deep learning for cell detection on highly imbalanced pathological data
• MGNiceNet: Unified Monocular Geometric Scene Understanding
• SemanticSugarBeets: A Multi-Task Framework and Dataset for Inspecting Harvest and Storage Characteristics of Sugar Beets
• A Dense Siamese U-Net trained with Edge Enhanced 3D IOU Loss for Image Co-segmentation
• Organ At Risk Segmentation with Multiple Modality
• A method for detecting text of arbitrary shapes in natural scenes that improves text spotting
• From Sleep Staging to Spindle Detection: A Case Study on End-to-End Automated Sleep Analysis
• Trustworthy AI-based crack-tip segmentation using domain-guided explanations
• Segmenting proto-halos with vision transformers
• Polynomial Dice Loss for Medical Image Segmentation
• Multi-lane Detection Using Instance Segmentation and Attentive Voting
• Physics-Informed CNN-LSTM for Street-Scale Urban Flood Prediction: Reconciling Aggregate Accuracy and Street-Level Plausibility