東華大學圖書館 |

Language: English

Help

回圖書館首頁

手機版館藏查詢

Back

Switch To: Labeled | MARC Mode | ISBD

Deep Learning With Limited Labeled D...

Huang, Zhe,

Linked to FindBook

Google Book

Amazon

博客來

Deep Learning With Limited Labeled Data: New Methods and Applications to Echocardiography /

Record Type:	Electronic resources : Monograph/item
Title/Author:	Deep Learning With Limited Labeled Data: New Methods and Applications to Echocardiography // Zhe Huang.
Author:	Huang, Zhe,
Description:	1 electronic resource (230 pages)
Notes:	Source: Dissertations Abstracts International, Volume: 86-03, Section: B.
Contained By:	Dissertations Abstracts International86-03B.
Subject:	Computer science. -
Online resource:	https://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=31556943
ISBN:	9798384086307

Deep Learning With Limited Labeled Data: New Methods and Applications to Echocardiography /
Huang, Zhe,

Deep Learning With Limited Labeled Data: New Methods and Applications to Echocardiography /Zhe Huang. - 1 electronic resource (230 pages)

Source: Dissertations Abstracts International, Volume: 86-03, Section: B.

Deep neural networks have revolutionized various fields via high-quality flexible predictions. However, their effectiveness often hinges on the availability of large labeled datasets for training-a limitation that is particularly pronounced in medical applications. This thesis addresses the challenge of learning from limited labeled data in three complementary areas:1. Dataset Development and Realistic Evaluation: We developed the Tufts Echocardiogram Datasets (TMED) (Huang et al., 2021b, 2022a). This dataset serves dual purposes: providing an authentic benchmark for semi-supervised learning (SSL) method development and encouraging effort toward a clinically impactful problem: diagnosing a heart disease known as Aortic Stenosis. In later work (Huang et al., 2024a), we developed a comprehensive benchmark of semi-supervised and self-supervised methods across four medical imaging tasks, helping practitioners with limited labeling and compute budget to achieve good results for tasks at hand.2.Method Development: We introduced two innovative approaches for semi-supervised learning. Fix-A-Step (Huang et al., 2023b) challenges traditional views on out-of-distribution unlabeled data and proposes a simple-to-implement learning procedure that leverages beneficial aspects of such data while guarding against harmful effects. InterLUDE (Huang et al., 2024c) addresses the disconnect between labeled and unlabeled data sets prevalent in current SSL formulations, proposing a novel approach that enhances SSL through beneficial interactions between labeled and unlabeled data.3.Clinical Translation: Focusing on the automated screening of Aortic Stenosis (AS), we conducted a proof-of-concept study (Wessler et al., 2023) demonstrating the feasibility of using deep learning for AS detection. Further studies improved the model by adopting a Multiple-Instance Learning (MIL) framework (Huang et al., 2023d) and incorporating multiple imaging modalities used in clinical practice (Huang et al., 2024b), significantly advancing AS diagnosis performance.

English

ISBN: 9798384086307Subjects--Topical Terms:

523869
Computer science.
Subjects--Index Terms:

Semi-supervised learning

Deep Learning With Limited Labeled Data: New Methods and Applications to Echocardiography /
LDR:03547nmm a22004453i 4500 001 2434689
005 20250923060941.5
006 m o d
007 cr|nu||||||||
008 261011s2024 miu||||||m |||||||eng d
020 $a 9798384086307
035 $a (MiAaPQD)AAI31556943
035 $a AAI31556943
040 $a MiAaPQD $b eng $c MiAaPQD $e rda
100 1 $a Huang, Zhe, $e author. $3 3824754
245 1 0 $a Deep Learning With Limited Labeled Data: New Methods and Applications to Echocardiography / $c Zhe Huang.
264 1 $a Ann Arbor : $b ProQuest Dissertations & Theses, $c 2024
300 $a 1 electronic resource (230 pages)
336 $a text $b txt $2 rdacontent
337 $a computer $b c $2 rdamedia
338 $a online resource $b cr $2 rdacarrier
500 $a Source: Dissertations Abstracts International, Volume: 86-03, Section: B.
500 $a Advisors: Hughes, Michael Committee members: Wessler, Benjamin; Jacob, Robert; Liu, Liping; Mortazavi, Bobak.
502 $b Ph.D. $c Tufts University $d 2024.
520 $a Deep neural networks have revolutionized various fields via high-quality flexible predictions. However, their effectiveness often hinges on the availability of large labeled datasets for training-a limitation that is particularly pronounced in medical applications. This thesis addresses the challenge of learning from limited labeled data in three complementary areas:1. Dataset Development and Realistic Evaluation: We developed the Tufts Echocardiogram Datasets (TMED) (Huang et al., 2021b, 2022a). This dataset serves dual purposes: providing an authentic benchmark for semi-supervised learning (SSL) method development and encouraging effort toward a clinically impactful problem: diagnosing a heart disease known as Aortic Stenosis. In later work (Huang et al., 2024a), we developed a comprehensive benchmark of semi-supervised and self-supervised methods across four medical imaging tasks, helping practitioners with limited labeling and compute budget to achieve good results for tasks at hand.2.Method Development: We introduced two innovative approaches for semi-supervised learning. Fix-A-Step (Huang et al., 2023b) challenges traditional views on out-of-distribution unlabeled data and proposes a simple-to-implement learning procedure that leverages beneficial aspects of such data while guarding against harmful effects. InterLUDE (Huang et al., 2024c) addresses the disconnect between labeled and unlabeled data sets prevalent in current SSL formulations, proposing a novel approach that enhances SSL through beneficial interactions between labeled and unlabeled data.3.Clinical Translation: Focusing on the automated screening of Aortic Stenosis (AS), we conducted a proof-of-concept study (Wessler et al., 2023) demonstrating the feasibility of using deep learning for AS detection. Further studies improved the model by adopting a Multiple-Instance Learning (MIL) framework (Huang et al., 2023d) and incorporating multiple imaging modalities used in clinical practice (Huang et al., 2024b), significantly advancing AS diagnosis performance.
546 $a English
590 $a School code: 0234
650 4 $a Computer science. $3 523869
650 4 $a Biomedical engineering. $3 535387
650 4 $a Medical imaging. $3 3172799
653 $a Semi-supervised learning
653 $a Deep neural networks
653 $a Heart disease
653 $a Echocardiogram
653 $a Aortic Stenosis
690 $a 0984
690 $a 0800
690 $a 0541
690 $a 0574
710 2 $a Tufts University. $b Computer Science. $e degree granting institution. $3 3816160
720 1 $a Hughes, Michael $e degree supervisor.
773 0 $t Dissertations Abstracts International $g 86-03B.
790 $a 0234
791 $a Ph.D.
792 $a 2024
856 4 0 $u https://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=31556943