A feature transfer enabled multi-task deep learning model on medical imaging

Fei Gao, Hyunsoo Yoon, Teresa Wu, Xianghua Chu

Research output: Contribution to journalArticle

Abstract

Object detection, segmentation, and classification are three common tasks in medical image analysis. Multi-task deep learning (MTL) tackles these three tasks jointly, which provides two advantages—saving computational cost and improving robustness against overfitting. Existing multi-task deep models start with learning each task as an individual objective in parallel and then integrate the tasks at the end of the architecture with one cost function. Such architecture fails to take advantage of the combined power of the features from each individual task at an early stage of the training. In this research, we propose a new architecture, FT-MTL-Net, an MTL model enabled by feature transfer. Traditional transfer learning deals with the same or similar task (e.g., classification) from different data sources (a.k.a. domain). The underlying assumption is that the knowledge gained from various source domains may help the learning task on the target domain. Our proposed FT-MTL-Net utilizes the different tasks from the same domain. Considering that features from the tasks are different views of the domain, the combined feature maps can be well exploited using knowledge from multiple views to enhance the generalizability. To evaluate the validity of the proposed approach, FT-MTL-Net is compared with models from literature including eight classification models, four detection models, and three segmentation models using a publicly available Full Filed Digital Mammogram dataset for breast cancer diagnosis. Experimental results show that the proposed FT-MTL-Net outperforms the competing models in classification and detection and has comparable results in segmentation.

Original languageEnglish (US)
Article number112957
JournalExpert Systems With Applications
Volume143
DOIs
StatePublished - Apr 1 2020

Fingerprint

Medical imaging
Deep learning
Cost functions
Image analysis
Costs

Keywords

  • Classification
  • Medical imaging analysis
  • Multi-task deep learning
  • Object detection
  • Segmentation

ASJC Scopus subject areas

  • Engineering(all)
  • Computer Science Applications
  • Artificial Intelligence

Cite this

A feature transfer enabled multi-task deep learning model on medical imaging. / Gao, Fei; Yoon, Hyunsoo; Wu, Teresa; Chu, Xianghua.

In: Expert Systems With Applications, Vol. 143, 112957, 01.04.2020.

Research output: Contribution to journalArticle

@article{48f31239aeee4e918445ace655d57245,
title = "A feature transfer enabled multi-task deep learning model on medical imaging",
abstract = "Object detection, segmentation, and classification are three common tasks in medical image analysis. Multi-task deep learning (MTL) tackles these three tasks jointly, which provides two advantages—saving computational cost and improving robustness against overfitting. Existing multi-task deep models start with learning each task as an individual objective in parallel and then integrate the tasks at the end of the architecture with one cost function. Such architecture fails to take advantage of the combined power of the features from each individual task at an early stage of the training. In this research, we propose a new architecture, FT-MTL-Net, an MTL model enabled by feature transfer. Traditional transfer learning deals with the same or similar task (e.g., classification) from different data sources (a.k.a. domain). The underlying assumption is that the knowledge gained from various source domains may help the learning task on the target domain. Our proposed FT-MTL-Net utilizes the different tasks from the same domain. Considering that features from the tasks are different views of the domain, the combined feature maps can be well exploited using knowledge from multiple views to enhance the generalizability. To evaluate the validity of the proposed approach, FT-MTL-Net is compared with models from literature including eight classification models, four detection models, and three segmentation models using a publicly available Full Filed Digital Mammogram dataset for breast cancer diagnosis. Experimental results show that the proposed FT-MTL-Net outperforms the competing models in classification and detection and has comparable results in segmentation.",
keywords = "Classification, Medical imaging analysis, Multi-task deep learning, Object detection, Segmentation",
author = "Fei Gao and Hyunsoo Yoon and Teresa Wu and Xianghua Chu",
year = "2020",
month = "4",
day = "1",
doi = "10.1016/j.eswa.2019.112957",
language = "English (US)",
volume = "143",
journal = "Expert Systems with Applications",
issn = "0957-4174",
publisher = "Elsevier Limited",

}

TY - JOUR

T1 - A feature transfer enabled multi-task deep learning model on medical imaging

AU - Gao, Fei

AU - Yoon, Hyunsoo

AU - Wu, Teresa

AU - Chu, Xianghua

PY - 2020/4/1

Y1 - 2020/4/1

N2 - Object detection, segmentation, and classification are three common tasks in medical image analysis. Multi-task deep learning (MTL) tackles these three tasks jointly, which provides two advantages—saving computational cost and improving robustness against overfitting. Existing multi-task deep models start with learning each task as an individual objective in parallel and then integrate the tasks at the end of the architecture with one cost function. Such architecture fails to take advantage of the combined power of the features from each individual task at an early stage of the training. In this research, we propose a new architecture, FT-MTL-Net, an MTL model enabled by feature transfer. Traditional transfer learning deals with the same or similar task (e.g., classification) from different data sources (a.k.a. domain). The underlying assumption is that the knowledge gained from various source domains may help the learning task on the target domain. Our proposed FT-MTL-Net utilizes the different tasks from the same domain. Considering that features from the tasks are different views of the domain, the combined feature maps can be well exploited using knowledge from multiple views to enhance the generalizability. To evaluate the validity of the proposed approach, FT-MTL-Net is compared with models from literature including eight classification models, four detection models, and three segmentation models using a publicly available Full Filed Digital Mammogram dataset for breast cancer diagnosis. Experimental results show that the proposed FT-MTL-Net outperforms the competing models in classification and detection and has comparable results in segmentation.

AB - Object detection, segmentation, and classification are three common tasks in medical image analysis. Multi-task deep learning (MTL) tackles these three tasks jointly, which provides two advantages—saving computational cost and improving robustness against overfitting. Existing multi-task deep models start with learning each task as an individual objective in parallel and then integrate the tasks at the end of the architecture with one cost function. Such architecture fails to take advantage of the combined power of the features from each individual task at an early stage of the training. In this research, we propose a new architecture, FT-MTL-Net, an MTL model enabled by feature transfer. Traditional transfer learning deals with the same or similar task (e.g., classification) from different data sources (a.k.a. domain). The underlying assumption is that the knowledge gained from various source domains may help the learning task on the target domain. Our proposed FT-MTL-Net utilizes the different tasks from the same domain. Considering that features from the tasks are different views of the domain, the combined feature maps can be well exploited using knowledge from multiple views to enhance the generalizability. To evaluate the validity of the proposed approach, FT-MTL-Net is compared with models from literature including eight classification models, four detection models, and three segmentation models using a publicly available Full Filed Digital Mammogram dataset for breast cancer diagnosis. Experimental results show that the proposed FT-MTL-Net outperforms the competing models in classification and detection and has comparable results in segmentation.

KW - Classification

KW - Medical imaging analysis

KW - Multi-task deep learning

KW - Object detection

KW - Segmentation

UR - http://www.scopus.com/inward/record.url?scp=85074329588&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85074329588&partnerID=8YFLogxK

U2 - 10.1016/j.eswa.2019.112957

DO - 10.1016/j.eswa.2019.112957

M3 - Article

AN - SCOPUS:85074329588

VL - 143

JO - Expert Systems with Applications

JF - Expert Systems with Applications

SN - 0957-4174

M1 - 112957

ER -