Chest X-Ray Bone Suppression for Improving Classification of\n Tuberculosis-Consistent Findings

2021/04/09 by Sivaramakrishnan Rajaraman, Rajaraman, Sivaramakrishnan, Ghada Zamzmi +7 · 4 citations
Medicine · #COVID-19 diagnosis using AI #Computer Vision and Pattern Recognition (cs.CV) #FOS: Computer and information sciences #FOS: Electrical engineering #I.4.5 #Image and Video Processing (eess.IV) #Infectious Diseases and Tuberculosis #Radiomics and Machine Learning in Medical Imaging #Tuberculosis Research and Epidemiology #electronic engineering #information engineering

paper · pdf · doi:10.48550/arxiv.2104.04518

openalex publication_date 2021/04/09 · openalex created_date 2022/07/25 · openalex updated_date 2026/07/28

Abstract

Chest X-rays are the most commonly performed diagnostic examination to detect\ncardiopulmonary abnormalities. However, the presence of bony structures such as\nribs and clavicles can obscure subtle abnormalities, resulting in diagnostic\nerrors. This study aims to build a deep learning-based bone suppression model\nthat identifies and removes these occluding bony structures in frontal CXRs to\nassist in reducing errors in radiological interpretation, including DL\nworkflows, related to detecting manifestations consistent with tuberculosis\n(TB). Several bone suppression models with various deep architectures are\ntrained and optimized using the proposed combined loss function and their\nperformances are evaluated in a cross-institutional test setting. The\nbest-performing model is used to suppress bones in the publicly available\nShenzhen and Montgomery TB CXR collections. A VGG-16 model is pretrained on a\nlarge collection of publicly available CXRs. The CXR-pretrained model is then\nfine-tuned individually on the non-bone-suppressed and bone-suppressed CXRs of\nShenzhen and Montgomery TB CXR collections to classify them as showing normal\nlungs or TB manifestations. The performances of these models are compared using\nseveral performance metrics, analyzed for statistical significance, and their\npredictions are qualitatively interpreted through class-selective relevance\nmaps. It is observed that the models trained on bone-suppressed CXRs\nsignificantly outperformed (p<0.05) the models trained on the\nnon-bone-suppressed CXRs. Models trained on bone-suppressed CXRs improved\ndetection of TB-consistent findings and resulted in compact clustering of the\ndata points in the feature space signifying that bone suppression improved the\nmodel sensitivity toward TB classification.\n

Citations

• Very Deep Convolutional Networks for Large-Scale Image Recognition
• Deep Convolutional Neural Network for Inverse Problems in Imaging

Cited by

• RepViT-CXR: A Channel Replication Strategy for Vision Transformers in Chest X-ray Tuberculosis and Pneumonia Classification
• GL-LCM: Global-Local Latent Consistency Models for Fast High-Resolution Bone Suppression in Chest X-Ray Images
• X-ray transferable polyrepresentation learning
• MedPrompt: LLM-CNN Fusion with Weight Routing for Medical Image Segmentation and Classification

Related

• Deep Learning for Automated Classification of Tuberculosis-Related Chest X-Ray: Dataset Specificity Limits Diagnostic Performance Generalizability
• Weakly Supervised Tuberculosis Localization in Chest X-rays through Knowledge Distillation
• Refining Tuberculosis Detection in CXR Imaging: Addressing Bias in Deep Neural Networks via Interpretability
• Tackling Tuberculosis: A Comparative Dive into Machine Learning for Tuberculosis Detection
• When does Bone Suppression and Lung Field Segmentation Improve Chest X-Ray Disease Classification?