Anatomy of Domain Shift Impact on U-Net Layers in MRI Segmentation

Zakazov, Ivan; Shirokikh, Boris; Chernyavskiy, Alexey; Belyaev, Mikhail

doi:10.1007/978-3-030-87199-4_20

Cited by 20 publications

(11 citation statements)

References 15 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…We also note that P-Consistency operates with the last decoder layer; thus, we compare it with F-Cons (Enc) and show it resulting in the significantly lower consistency score, 0.63 against 0.80 (p 1 < 0.05, p 2 < 10 −10 ). Thus, our experimental evidences align with the message of (Zakazov et al, 2021) that the earlier (encoder) layers contain more domain-specific information than the later (decoder, output) ones.…”

Section: Resultssupporting

confidence: 83%

See 1 more Smart Citation

Adaptation to CT Reconstruction Kernels by Enforcing Cross-domain Feature Maps Consistency

Shimovolos¹,

Shushko²,

Belyaev³

et al. 2022

Preprint

Self Cite

View full text Add to dashboard Cite

Deep learning methods provide significant assistance in analyzing coronavirus disease in chest computed tomography (CT) images, including identification, severity assessment, and segmentation. Although the earlier developed methods address the lack of data and specific annotations, the current goal is to build a robust algorithm for clinical use, having a larger pool of available data. With the larger datasets, the domain shift problem arises, affecting the performance of methods on the unseen data. One of the critical sources of domain shift in CT images is the difference in reconstruction kernels used to generate images from the raw data (sinograms). In this paper, we show a decrease in the COVID-19 segmentation quality of the model trained on the smooth and tested on the sharp reconstruction kernels. Furthermore, we compare several domain adaptation approaches to tackle the problem, such as task-specific augmentation and unsupervised adversarial learning. Finally, we propose the unsupervised adaptation method, called F-Consistency, that outperforms the previous approaches. Our method exploits a set of unlabeled CT image pairs which differ only in reconstruction kernels within every pair. It enforces the similarity of the network's hidden representations (feature maps) by minimizing mean squared error (MSE) between paired feature maps. We show our method achieving 0.64 Dice Score on the test dataset with unseen sharp kernels, compared to the 0.56 Dice Score of the baseline model. Moreover, F-Consistency scores 0.80 Dice Score between predictions on the paired images, which almost doubles the baseline score of 0.46 and surpasses the other methods. We also show F-Consistency to better generalize on the unseen kernels and without the specific semantic content, e.g., presence of the COVID-19 lesions.

show abstract

Section: Resultssupporting

confidence: 83%

“…However, there is no consensus in the literature on how to connect the discriminator to a segmentation network (Zakazov et al, 2021). Our experiments also show a high dependency of the model performance from the connection implementation.…”

Section: Deep Adversarial Neural Networkmentioning

confidence: 73%

Adaptation to CT Reconstruction Kernels by Enforcing Cross-domain Feature Maps Consistency

Shimovolos¹,

Shushko²,

Belyaev³

et al. 2022

Preprint

Self Cite

View full text Add to dashboard Cite

show abstract

“…Magnetic Resonance Imaging (MRI) data is even more susceptible to changes in the acquisition conditions than CTs, as there is no consensus on the calibration of intensity values. This causes the performance of segmentation models trained on MR tasks to deteriorate on OOD data ( Zakazov et al, 2021 , Kondrateva et al, 2021 ).…”

Section: Resultsmentioning

confidence: 99%

Distance-based detection of out-of-distribution silent failures for Covid-19 lung lesion segmentation

González

Gotkowski

Fuchs

et al. 2022

Medical Image Analysis

View full text Add to dashboard Cite

“…However, shifts in data statistics, i.e. domain shifts [41,42], can heavily degrade deep model performance on unseen data [4,24,33]. By contrast, humans can learn quickly with limited supervision and are less likely to be affected by such domain shifts, achieving accurate recognition on new images from different 𝑅 ( :…”

Section: Introductionmentioning

confidence: 99%

vMFNet: Compositionality Meets Domain-Generalised Segmentation

Liu¹,

Thermos²,

Sánchez³

et al. 2022

Lecture Notes in Computer Science

View full text Add to dashboard Cite

Training medical image segmentation models usually requires a large amount of labeled data. By contrast, humans can quickly learn to accurately recognise anatomy of interest from medical (e.g. MRI and CT) images with some limited guidance. Such recognition ability can easily generalise to new images from different clinical centres. This rapid and generalisable learning ability is mostly due to the compositional structure of image patterns in the human brain, which is less incorporated in medical image segmentation. In this paper, we model the compositional components (i.e. patterns) of human anatomy as learnable von-Mises-Fisher (vMF) kernels, which are robust to images collected from different domains (e.g. clinical centres). The image features can be decomposed to (or composed by) the components with the composing operations, i.e. the vMF likelihoods. The vMF likelihoods tell how likely each anatomical part is at each position of the image. Hence, the segmentation mask can be predicted based on the vMF likelihoods. Moreover, with a reconstruction module, unlabeled data can also be used to learn the vMF kernels and likelihoods by recombining them to reconstruct the input image. Extensive experiments show that the proposed vMFNet achieves improved generalisation performance on two benchmarks, especially when annotations are limited. Code is publicly available at: https://github.com/vios-s/vMFNet.

show abstract

Anatomy of Domain Shift Impact on U-Net Layers in MRI Segmentation

Cited by 20 publications

References 15 publications

Adaptation to CT Reconstruction Kernels by Enforcing Cross-domain Feature Maps Consistency

Adaptation to CT Reconstruction Kernels by Enforcing Cross-domain Feature Maps Consistency

Distance-based detection of out-of-distribution silent failures for Covid-19 lung lesion segmentation

vMFNet: Compositionality Meets Domain-Generalised Segmentation

Contact Info

Product

Resources

About