Electrical Engineering and Systems Science > Image and Video Processing
[Submitted on 31 Aug 2024]
Title:Statistical Distance-Guided Unsupervised Domain Adaptation for Automated Multi-Class Cardiovascular Magnetic Resonance Image Quality Assessment
View PDFAbstract:This study proposes an attention-based statistical distance-guided unsupervised domain adaptation model for multi-class cardiovascular magnetic resonance (CMR) image quality assessment. The proposed model consists of a feature extractor, a label predictor and a statistical distance estimator. An annotated dataset as the source set and an unlabeled dataset as the target set with different statistical distributions are considered inputs. The statistical distance estimator approximates the Wasserstein distance between the extracted feature vectors from the source and target data in a mini-batch. The label predictor predicts data labels of source data and uses a combinational loss function for training, which includes cross entropy and centre loss functions plus the estimated value of the distance estimator. Four datasets, including imaging and k-space data, were used to evaluate the proposed model in identifying four common CMR imaging artefacts: respiratory and cardiac motions, Gibbs ringing and Aliasing. The results of the extensive experiments showed that the proposed model, both in image and k-space analysis, has an acceptable performance in covering the domain shift between the source and target sets. The model explainability evaluations and the ablation studies confirmed the proper functioning and effectiveness of all the model's modules. The proposed model outperformed the previous studies regarding performance and the number of examined artefacts. The proposed model can be used for CMR post-imaging quality control or in large-scale cohort studies for image and k-space quality assessment due to the appropriate performance in domain shift coverage without a tedious data-labelling process.
Submission history
From: Shahabedin Nabavi [view email][v1] Sat, 31 Aug 2024 07:50:07 UTC (1,448 KB)
References & Citations
Bibliographic and Citation Tools
Bibliographic Explorer (What is the Explorer?)
Connected Papers (What is Connected Papers?)
Litmaps (What is Litmaps?)
scite Smart Citations (What are Smart Citations?)
Code, Data and Media Associated with this Article
alphaXiv (What is alphaXiv?)
CatalyzeX Code Finder for Papers (What is CatalyzeX?)
DagsHub (What is DagsHub?)
Gotit.pub (What is GotitPub?)
Hugging Face (What is Huggingface?)
Papers with Code (What is Papers with Code?)
ScienceCast (What is ScienceCast?)
Demos
Recommenders and Search Tools
Influence Flower (What are Influence Flowers?)
CORE Recommender (What is CORE?)
arXivLabs: experimental projects with community collaborators
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.