Incremental Learning Meets Transfer Learning: Application to Multi-site Prostate MRI Segmentation

  title={Incremental Learning Meets Transfer Learning: Application to Multi-site Prostate MRI Segmentation},
  author={Chenyu You and Jinlin Xiang and Kun Su and Xiaoran Zhang and Siyuan Dong and John A. Onofrey and Lawrence H. Staib and James S. Duncan},
. Many medical datasets have recently been created for medical image segmentation tasks, and it is natural to question whether we can use them to sequentially train a single model that (1) performs better on all these datasets, and (2) generalizes well and transfers better to the unknown target site domain. Prior works have achieved this goal by jointly training one model on multi-site datasets, which achieve competitive performance on average but such methods rely on the assumption about the… 

Figures and Tables from this paper

Bootstrapping Semi-supervised Medical Image Segmentation with Anatomical-aware Contrastive Distillation

The key innovation in ACTION is to learn global semantic relationship across the entire dataset and local anatomical features among the neighbouring pixels with minimal additional memory footprint, which can generate smoother segmentation boundaries and more accurate predictions.

TKIL: Tangent Kernel Approach for Class Balanced Incremental Learning

This work proposes the introduction of a novel methodology of Tangent Kernel for Incremental Learning (TKIL) that achieves class-balanced performance, and shows that TKIL outperforms existing state-of-the-art methods on various incremental learning settings.

Mine yOur owN Anatomy: Revisiting Medical Image Segmentation with Extremely Limited Labels

This paper introduces a novel semi-supervised 2D medical image segmentation framework termed MONA, and constructs a set of objectives that encourage the model to be capable of decomposing medical images into a collection of anatomical features in an unsupervised manner.

Learning correspondences of cardiac motion from images using biomechanics-informed modeling

This work proposes an explicit biomechanics-informed prior as regularization on the pre-dicted DVF in modeling a more generic biomechanically plausible trans-formation within all cardiac structures without introducing additional training complexity.

Diffeomorphic Image Registration with Neural Velocity Field

A new optimization-based method named DNVF (Diffeomorphic Image Registration with Neural Velocity Field) which utilizes deep neural network to model the space of admissible transformations, and a cascaded image registration framework (Cas-DNVF) by combining the benefits of both optimization and learning based methods.

Class-Aware Adversarial Transformers for Medical Image Segmentation

CASTformer is presented, a novel type of adversarial transformers, for 2D medical image segmentation that takes advantage of the pyramid structure to construct multi- scale representations and handle multi-scale variations, and designs a novel class-aware transformer module to better learn the discriminative regions of objects with semantic structures.



MS-Net: Multi-Site Network for Improving Prostate Segmentation With Heterogeneous MRI Data

A novel multi-site network (MS-Net) is proposed for improving prostate segmentation by learning robust representations, leveraging multiple sources of data and develops Domain-Specific Batch Normalization layers in the network backbone to compensate for inter-site heterogeneity of different MRI datasets.

Shape-aware Meta-learning for Generalizing Prostate MRI Segmentation to Unseen Domains

A novel shape-aware meta-learning scheme is presented to improve the model generalization in prostate MRI segmentation, and outperforms many state-of-the-art generalization methods consistently across all six settings of unseen domains.

Incremental Learning for Multi-organ Segmentation with Partially Labeled Datasets

This work gives the first attempt to conjecture that the different distribution is the key reason for ‘catastrophic forgetting’ that commonly exists in IL methods, and verify that IL has the natural adaptability to medical image scenarios.

HD-Net: Hybrid Discriminative Network for Prostate Segmentation in MR Images

The Hybrid Discriminative Network (HD-Net), which consists of a 3D segmentation decoder using channel attention block to generate semantically consistent volumetric features and an auxiliary 2D boundary decoder guiding the segmentation network to focus on the semantically discriminative intra-slice features, is proposed.

Unpaired Multi-Modal Segmentation via Knowledge Distillation

This work proposes a novel learning scheme for unpaired cross-modality image segmentation, with a highly compact architecture achieving superior segmentation accuracy, and introduces a novel loss term inspired by knowledge distillation.

Inter-site Variability in Prostate Segmentation Accuracy Using Deep Learning

It is found that the segmentation accuracy of a single-site network was substantially worse on data from unseen sites than on dataFrom the training site, and two strategies for mitigating the reduced accuracy are evaluated.

ASDNet: Attention Based Semi-supervised Deep Networks for Medical Image Segmentation

A fully convolutional confidence network is proposed to adversarially train the segmentation network and a region-attention based semi-supervised learning strategy to include the unlabeled data for training is proposed.

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

This work proposes an approach to 3D image segmentation based on a volumetric, fully convolutional, neural network, trained end-to-end on MRI volumes depicting prostate, and learns to predict segmentation for the whole volume at once.

A Lifelong Learning Approach to Brain MR Segmentation Across Scanners and Protocols

This work proposes a single CNN with shared convolutional filters and domain-specific batch normalization layers, which can be tuned to new domains with only a few (\({\approx 4) labelled images), which largely closes the gap to the benchmark, which is training a dedicated CNN for each scanner.

Marginal loss and exclusion loss for partially supervised multi-organ segmentation