Lung image segmentation with improved U-Net, V-Net and Seg-Net techniques.

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Computer-Aided Diagnosis

Background:

• Tuberculosis (TB) is a major global health concern requiring accurate diagnostic tools.
• Image analysis using machine learning is emerging as a key method for TB detection.
• Accurate segmentation of lung regions is vital for diagnosing respiratory diseases like TB.

Purpose of the Study:

• To develop and evaluate deep learning models for enhanced tuberculosis segmentation in medical images.
• To improve the accuracy and reliability of tuberculosis detection through advanced image analysis techniques.

Main Methods:

• Proposed three segmentation models: U-Net, V-Net, and Seg-Net architectures.
• Utilized Shenzhen and Montgomery chest X-ray databases for training and validation.
• Incorporated advanced preprocessing, attention mechanisms, and non-local blocks to boost segmentation performance.

Main Results:

• U-Net and V-Net models achieved high Dice coefficient scores (96.43% and 96.42%, respectively).
• The proposed deep learning models outperformed traditional methods in segmentation accuracy.
• Demonstrated robust performance and reliable lung region segmentation on benchmark datasets.

Conclusions:

• The developed U-Net and V-Net models show superior effectiveness for tuberculosis segmentation compared to Seg-Net and traditional approaches.
• Accurate lung segmentation using these AI models is crucial for precise diagnosis of tuberculosis and other respiratory conditions.
• These findings highlight the potential of deep learning in advancing medical image analysis for disease detection.