Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Enhanced Tuberculosis Bacilli Detection using Attention-Residual U-Net and Ensemble Classification

Greeshma K, Vishnukumar S

TL;DR

The paper addresses automated detection of tuberculosis bacilli from bright-field sputum smear images, where prior methods often suffer from limited automation and segmentation/classification performance. It presents a hybrid two-stage approach: an Attention Residual U-Net for precise segmentation to extract Regions of Interest, followed by an ensemble classifier (SVM, Random Forest, XGBoost) for bacilli identification. The authors introduce the DCA-CUSAT TB Dataset and validate the method on this and additional public datasets (Costa, ZNSM-iDB), reporting superior segmentation metrics and higher classification accuracy relative to prior works. The work demonstrates significant potential for automated, accurate TB screening and could reduce manual workload in microscopy workflows.

Abstract

Tuberculosis (TB), caused by Mycobacterium tuberculosis, remains a critical global health issue, necessitating timely diagnosis and treatment. Current methods for detecting tuberculosis bacilli from bright field microscopic sputum smear images suffer from low automation, inadequate segmentation performance, and limited classification accuracy. This paper proposes an efficient hybrid approach that combines deep learning for segmentation and an ensemble model for classification. An enhanced U-Net model incorporating attention blocks and residual connections is introduced to precisely segment microscopic sputum smear images, facilitating the extraction of Regions of Interest (ROIs). These ROIs are subsequently classified using an ensemble classifier comprising Support Vector Machine (SVM), Random Forest, and Extreme Gradient Boost (XGBoost), resulting in an accurate identification of bacilli within the images. Experiments conducted on a newly created dataset, along with public datasets, demonstrate that the proposed model achieves superior segmentation performance, higher classification accuracy, and enhanced automation compared to existing methods.

Enhanced Tuberculosis Bacilli Detection using Attention-Residual U-Net and Ensemble Classification

TL;DR

The paper addresses automated detection of tuberculosis bacilli from bright-field sputum smear images, where prior methods often suffer from limited automation and segmentation/classification performance. It presents a hybrid two-stage approach: an Attention Residual U-Net for precise segmentation to extract Regions of Interest, followed by an ensemble classifier (SVM, Random Forest, XGBoost) for bacilli identification. The authors introduce the DCA-CUSAT TB Dataset and validate the method on this and additional public datasets (Costa, ZNSM-iDB), reporting superior segmentation metrics and higher classification accuracy relative to prior works. The work demonstrates significant potential for automated, accurate TB screening and could reduce manual workload in microscopy workflows.

Abstract

Tuberculosis (TB), caused by Mycobacterium tuberculosis, remains a critical global health issue, necessitating timely diagnosis and treatment. Current methods for detecting tuberculosis bacilli from bright field microscopic sputum smear images suffer from low automation, inadequate segmentation performance, and limited classification accuracy. This paper proposes an efficient hybrid approach that combines deep learning for segmentation and an ensemble model for classification. An enhanced U-Net model incorporating attention blocks and residual connections is introduced to precisely segment microscopic sputum smear images, facilitating the extraction of Regions of Interest (ROIs). These ROIs are subsequently classified using an ensemble classifier comprising Support Vector Machine (SVM), Random Forest, and Extreme Gradient Boost (XGBoost), resulting in an accurate identification of bacilli within the images. Experiments conducted on a newly created dataset, along with public datasets, demonstrate that the proposed model achieves superior segmentation performance, higher classification accuracy, and enhanced automation compared to existing methods.
Paper Structure (9 sections, 6 equations, 9 figures, 7 tables)

This paper contains 9 sections, 6 equations, 9 figures, 7 tables.

Table of Contents

  1. Introduction
  2. Related Works
  3. Methodology
  4. Experimental Results and Analysis
  5. DCA-CUSAT Bright Field Microscopic Sputum Smear TB Dataset
  6. Performance Analysis of the Proposed Model Using DCA-CUSAT TB dataset
  7. Performance Analysis of the Proposed Model Using Costa Dataset
  8. Performance Analysis of the Proposed Model Using ZNSM-iDB Dataset
  9. Conclusion

Figures (9)

  • Figure 1: Proposed System Architecture.
  • Figure 2: Proposed Attention Residual U-Net for segmentation.
  • Figure 3: Architecture of the Residual Convolution.
  • Figure 4: Nikon Ti2-u Eclipse microscope, integrated with the NIS-elements software package at Department of Biotechnology, Cochin University of Science and Technology.
  • Figure 5: Sample images - DCA-CUSAT Bright Field Microscopic Sputum Smear TB Dataset.
  • ...and 4 more figures