Abstract:Mycobacterium Tuberculosis (TB) is an infectious bacterial disease. In 2018, about 10 million people has been diagnosed with tuberculosis (TB) worldwide. Early diagnosis of TB is necessary for effective treatment, higher survival rate, and preventing its further transmission. The gold standard for tuberculosis diagnosis is sputum culture. Nevertheless, posterior-anterior chest radiographs (CXR) is an effective central method with low cost and a relatively low radiation dose for screening TB with immediate results. TB diagnosis from CXR is a challenging task requiring high level of expertise due to the diverse presentation of the disease. Significant intra-class variation and inter-class similarity in CXR images makes TB diagnosis from CXR a more challenging task. The main aim of this study is tuberculosis recognition from CXR images for reducing the disease burden. For this purpose, a novel multi-instance classification model is proposed in this study which is based on CNNs, complex networks and stacked ensemble (CCNSE). A main advantage of CCNSE is not requiring an accurate lung segmentation to localize the suspicious regions. Several overlapping patches are extracted from each CXR image. Features describing each patch are obtained by CNNs and then the feature vectors are clustered. Local complex networks (LCN) and global ones (GCN) of the cluster representatives are formed and feature engineering on LCN (GCN) generates other features at image-level (patch-level and image-level). Global clustering on these feature sets is performed for all patches. Each patch is assigned the purity score of its corresponding cluster. Patch-level features and purity scores are aggregated for each image. Finally, the images are classified with a proposed stacked ensemble classifier to normal and TB classes. Two datasets are used in this study including Montgomery County CXR set (MC) and Shenzhen dataset (SZ). MC/SZ includes 138/662 chest X-rays (CXR) from which 80 and 58/326 and 336 images belong to normal/TB classes, respectively. The experimental results show that the proposed method with AUC of 99.00 ± 0.28/98.00 ± 0.16 for MC/SZ and accuracy of 99.26 ± 0.40/99.22 ± 0.32 for MC/SZ with fivefold cross validation strategy is superior than the compared ones for diagnosis of TB from CXR images. The proposed method can be used as a computer-aided diagnosis system to reduce the manual time, effort and dependency to specialist's expertise level.

Modality-Specific Deep Learning Model Ensembles Toward Improving TB Detection in Chest Radiographs

Detecting Tuberculosis-Consistent Findings in Lateral Chest X-Rays Using an Ensemble of CNNs and Vision Transformers

A robust ensemble deep learning framework for accurate diagnoses of tuberculosis from chest radiographs

Deep learning-based automatic detection of tuberculosis disease in chest X-ray images

Generalizable disease detection using model ensemble on chest X-ray images

Proposing a novel multi-instance learning model for tuberculosis recognition from chest X-ray images based on CNNs, complex networks and stacked ensemble

Improving Tuberculosis Detection in Chest X-ray Images through Transfer Learning and Deep Learning: A Comparative Study of CNN Architectures

Multifaceted Disease Diagnosis: Leveraging Transfer Learning with Deep Convolutional Neural Networks on Chest X-Rays for COVID-19, Pneumonia, and Tuberculosis

Deep Learning for Automated Screening of Tuberculosis from Indian Chest X-rays: Analysis and Update

SynthEnsemble: A Fusion of CNN, Vision Transformer, and Hybrid Models for Multi-Label Chest X-Ray Classification

Deep ensemble learning for segmenting tuberculosis-consistent manifestations in chest radiographs

Exploiting Cascaded Ensemble of Features for the Detection of Tuberculosis Using Chest Radiographs

Development and Validation of a 3D-Convolutional Neural Network Model Based on Chest CT for Differentiating Active Pulmonary Tuberculosis from Community-Acquired Pneumonia

Diagnosis of Pediatric Pneumonia with Ensemble of Deep Convolutional Neural Networks in Chest X-Ray Images

Ensembles of Convolutional Neural Networks models for pediatric pneumonia diagnosis

Inverted bell-curve-based ensemble of deep learning models for detection of COVID-19 from chest X-rays

Artificial Intelligence-based Deep Learning Architecture for Tuberculosis Detection

Iteratively Pruned Deep Learning Ensembles for COVID-19 Detection in Chest X-rays

Self-Trained Convolutional Neural Network (CNN) for Tuberculosis Diagnosis in Medical Imaging

Improved Semantic Segmentation of Tuberculosis-consistent findings in Chest X-rays Using Augmented Training of Modality-specific U-Net Models with Weak Localizations

A novel dense-net deep neural network with enhanced feature selection method for classification of different stages of tuberculosis using chest X-ray images