Search research articles

ABOUT JoVE

Overview Leadership Blog JoVE Help Center

AUTHORS

Publishing Process Editorial Board Scope & Policies Peer Review FAQ Submit

LIBRARIANS

Testimonials Subscriptions Access Resources Library Advisory Board FAQ

RESEARCH

JoVE Journal Methods Collections JoVE Encyclopedia of Experiments Archive

EDUCATION

JoVE Core JoVE Business JoVE Science Education JoVE Lab Manual Faculty Resource Center Faculty Site

Terms & Conditions of Use

Related Concept Videos

You might also read

Related Articles

Articles linked to this work by shared authors, journal, and citation graph.

Sort by

Same author

Considerations for Pre-deployment Planning in Point-of-Care Ultrasound Program Implementation: A HIMSS-SIIM Enterprise Imaging Community Whitepaper in Collaboration with AIUM.

Journal of imaging informatics in medicine·2026

Same author

Multimodal Learning with Privileged Report Supervision for Generalizable Tuberculosis Detection on Chest Radiographs.

Journal of medical systems·2026

Same author

Oral Cancer Detection By Using Tabular Data Synthesis and Classification.

Proceedings ... ICDM workshops. IEEE International Conference on Data Mining·2026

Same author

Artificial Intelligence-Based Diagnosis of Kaposi Sarcoma Using Digital Photographs in Dark-Skinned Patients in Uganda.

JCO global oncology·2026

Same author

Detecting Oral Cancer Using Tabular Deep Learning.

IEEE International Conference on Omni-layer Intelligent Systems : COINS. IEEE International Conference on Omni-layer Intelligent Systems·2025

Same author

Artificial Intelligence-based Diagnosis of Kaposi Sarcoma using Photographs in Dark-skinned Patients.

medRxiv : the preprint server for health sciences·2025

Search research articles

Related Experiment Video

Updated: Nov 10, 2025

Objectification of Tongue Diagnosis in Traditional Medicine, Data Analysis, and Study Application

Objectification of Tongue Diagnosis in Traditional Medicine, Data Analysis, and Study Application

Published on: April 14, 2023

Improved Semantic Segmentation of Tuberculosis-Consistent Findings in Chest X-rays Using Augmented Training of

Sivaramakrishnan Rajaraman¹, Les R Folio², Jane Dimperio²

¹National Library of Medicine, National Institutes of Health, Bethesda, MD 20894, USA.

Diagnostics (Basel, Switzerland)

|April 3, 2021

Summary

This summary is machine-generated.

Chest X-ray modality-specific U-Nets improve tuberculosis detection. Augmenting training data with weak localizations enhances segmentation accuracy for tuberculosis-consistent findings in medical images.

Keywords:

U-Net augmentation chest-X-rays convolutional neural networks deep learning localization modality-specific knowledge transfer segmentation tuberculosis

More Related Videos

Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography

Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography

Published on: November 30, 2022

Deep Learning-Based Segmentation of Cryo-Electron Tomograms

Deep Learning-Based Segmentation of Cryo-Electron Tomograms

Published on: November 11, 2022

Related Experiment Videos

Last Updated: Nov 10, 2025

Objectification of Tongue Diagnosis in Traditional Medicine, Data Analysis, and Study Application

Objectification of Tongue Diagnosis in Traditional Medicine, Data Analysis, and Study Application

Published on: April 14, 2023

Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography

Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography

Published on: November 30, 2022

Deep Learning-Based Segmentation of Cryo-Electron Tomograms

Deep Learning-Based Segmentation of Cryo-Electron Tomograms

Published on: November 11, 2022

Area of Science:

Medical Imaging
Artificial Intelligence
Radiology

Background:

Deep learning (DL) excels in image analysis, outperforming traditional methods in object localization and recognition.
U-Net segmentation models show superior performance in medical image analysis.
Modality-specific DL models enhance domain knowledge transfer, improving model adaptation and generalization for tasks like region of interest (ROI) localization.

Purpose of the Study:

To develop and evaluate chest X-ray (CXR) modality-specific U-Net models for semantic segmentation of tuberculosis (TB)-consistent findings.
To investigate the impact of data augmentation using weak localization on segmentation performance.
To automate the segmentation of TB manifestations to aid radiologists in diagnosis and improve patient care.

Main Methods:

Trained CXR modality-specific U-Net models and other state-of-the-art U-Net models on the TBX11K CXR dataset with weak TB annotations (bounding boxes).
Augmented training data with weak localization masks derived from a DL classifier trained to detect TB.
Evaluated model performance on test data from TBX11K and cross-institutional datasets (Shenzhen TB, Montgomery TB CXR).

Main Results:

CXR modality-specific U-Net models trained with the augmented strategy achieved superior performance.
The augmented training approach demonstrated significant improvements (p < 0.05) on both internal and external validation datasets.
The study is the first to utilize CXR modality-specific U-Nets for TB ROI segmentation and evaluate augmented training strategies.

Conclusions:

Augmenting training data with weak localizations significantly enhances the performance of CXR modality-specific U-Net models for TB detection.
The developed approach shows promise for improving the accuracy and efficiency of TB diagnosis through automated semantic segmentation.
This work establishes a novel methodology for training DL models in medical imaging, particularly for identifying specific pathologies like TB.