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

You, Chenyu; Xiang, Jinlin; Su, Kun; Zhang, Xiaoran; Dong, Siyuan; Onofrey, John A.; Staib, Lawrence H.; Duncan, James S.

doi:10.1007/978-3-031-18523-6_1

Cited by 19 publications

(3 citation statements)

References 37 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Domain adaptation explicitly seeks to bridge a given domain gap between a source domain with labelled data, and a specific target domain without labels. A first solution is to map both domains to a common latent space, where a classifier can be trained ( Kamnitsas et al, 2017a ; Dou et al, 2019 ; Ganin et al, 2017 ; You et al, 2022a ). In comparison, generative adaptation methods seek to match the source images to the target domain with image-to-image translation methods ( Sandfort et al, 2019 ; Huo et al, 2019 ; Zhang et al, 2018 ).…”

Section: Related Workmentioning

confidence: 99%

SynthSeg: Segmentation of brain MRI scans of any contrast and resolution without retraining

Billot

Greve

Puonti

et al. 2023

Medical Image Analysis

View full text Add to dashboard Cite

Section: Related Workmentioning

confidence: 99%

SynthSeg: Segmentation of brain MRI scans of any contrast and resolution without retraining

Billot

Greve

Puonti

et al. 2023

Medical Image Analysis

View full text Add to dashboard Cite

“…When faced with such words, the semantic information that the model can extract is weak, which leads to incorrect predictions. Therefore, it is imperative to incorporate text containing such words for the incremental training of our model [28] to rectify the semantic information of words. This approach can enhance the algorithm performance and continuously optimize the model in real-world business scenarios.…”

Section: Usability Verification Experiments For Real Business Scenariosmentioning

confidence: 99%

Automated Identification of Sensitive Financial Data Based on the Topic Analysis

Li,

Liu,

Yang

2024

Future Internet

View full text Add to dashboard Cite

Data governance is an extremely important protection and management measure throughout the entire life cycle of data. However, there are still data governance issues, such as data security risks, data privacy breaches, and difficulties in data management and access control. These problems lead to a risk of data breaches and abuse. Therefore, the security classification and grading of data has become an important task to accurately identify sensitive data and adopt appropriate maintenance and management measures with different sensitivity levels. This work started from the problems existing in the current data security classification and grading work, such as inconsistent classification and grading standards, difficult data acquisition and sorting, and weak semantic information of data fields, to find the limitations of the current methods and the direction for improvement. The automatic identification method of sensitive financial data proposed in this paper is based on topic analysis and was constructed by incorporating Jieba word segmentation, word frequency statistics, the skip-gram model, K-means clustering, and other technologies. Expert assistance was sought to select appropriate keywords for enhanced accuracy. This work used the descriptive text library and real business data of a Chinese financial institution for training and testing to further demonstrate its effectiveness and usefulness. The evaluation indicators illustrated the effectiveness of this method in the classification of data security. The proposed method addressed the challenge of sensitivity level division in texts with limited semantic information, which overcame the limitations on model expansion across different domains and provided an optimized application model. All of the above pointed out the direction for the real-time updating of the method.

show abstract

“…10 At the same time, the powerful parallel computing capability also enables the deep learning model to obtain good computing accuracy at a small time cost. On this basis, deep learning methods have been widely used in not only traditional X-CT segmentation and registration, [11][12][13][14] but also some novel medical imaging problems, such as sparse X-CT reconstruction [15][16][17] and dual-energy CT reconstruction. 18 On the other hand, the above advantages also become the main motivation for developing deep-learning-based medical image reconstruction methods, especially for LDCTs reconstruction.…”

Section: Introductionmentioning

confidence: 99%

Transfer learning framework for low‐dose CT reconstruction based on marginal distribution adaptation in multiscale

et al. 2023

View full text Add to dashboard Cite

Background: With the increasing use of computed tomography (CT) in clinical practice, limiting CT radiation exposure to reduce potential cancer risks has become one of the important directions of medical imaging research. As the dose decreases, the reconstructed CT image will be severely degraded by projection noise. Purpose: As an important method of image processing, supervised deep learning has been widely used in the restoration of low-dose CT (LDCT) in recent years. However, the normal-dose CT (NDCT) corresponding to a specific LDCT (it is regarded as the label of the LDCT, which is necessary for supervised learning) is very difficult to obtain so that the application of supervised learning methods in LDCT reconstruction is limited. It is necessary to construct a unsupervised deep learning framework for LDCT reconstruction that does not depend on paired LDCT-NDCT datasets. Methods: We presented an unsupervised learning framework for the transferring from the identity mapping to the low-dose reconstruction task, called marginal distribution adaptation in multiscale (MDAM). For NDCTs as source domain data, MDAM is an identity map with two parts: firstly, it establishes a dimensionality reduction mapping, which can obtain the same feature distribution from NDCTs and LDCTs; and then NDCTs is retrieved by reconstructing the image overview and details from the low-dimensional features. For the purpose of the feature transfer between source domain and target domain (LDCTs), we introduce the multiscale feature extraction in the MDAM, and then eliminate differences in probability distributions of these multiscale features between NDCTs and LDCTs through wavelet decomposition and domain adaptation learning. Results: Image quality evaluation metrics and subjective quality scores show that, as an unsupervised method, the performance of the MDAM approaches or even surpasses some state-of -the-art supervised methods. Especially, MDAM has been favorably evaluated in terms of noise suppression, structural preservation, and lesion detection. Conclusions: We demonstrated that, the MDAM framework can reconstruct corresponding NDCTs from LDCTs with high accuracy, and without relying on any labeles. Moreover, it is more suitable for clinical application compared with supervised learning methods.

show abstract

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

Cited by 19 publications

References 37 publications

SynthSeg: Segmentation of brain MRI scans of any contrast and resolution without retraining

SynthSeg: Segmentation of brain MRI scans of any contrast and resolution without retraining

Automated Identification of Sensitive Financial Data Based on the Topic Analysis

Transfer learning framework for low‐dose CT reconstruction based on marginal distribution adaptation in multiscale

Contact Info

Product

Resources

About