Secondary Pulmonary Tuberculosis Identification Via pseudo-Zernike Moment and Deep Stacked Sparse Autoencoder.

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Infectious Diseases

Background:

• Secondary pulmonary tuberculosis (SPT) is a leading cause of infectious disease mortality.
• Accurate and early detection of SPT is crucial for effective treatment and public health.

Purpose of the Study:

• To develop and evaluate a novel artificial intelligence model for improved recognition of secondary pulmonary tuberculosis.
• To enhance the diagnostic accuracy of SPT through advanced feature extraction and classification techniques.

Main Methods:

• A novel PZM-DSSAE model was proposed, utilizing Pseudo Zernike moments (PZM) for feature extraction and a deep stacked sparse autoencoder (DSSAE) for classification.
• 18-way data augmentation was implemented to mitigate overfitting and improve model robustness.
• The model's performance was rigorously assessed using 10-fold cross-validation over ten runs.

Main Results:

• The PZM-DSSAE model demonstrated high diagnostic performance, achieving a sensitivity of 93.33%, specificity of 93.13%, precision of 93.15%, accuracy of 93.23%, and F1 score of 93.23%.
• The model achieved an impressive area under the curve (AUC) of 0.9739.
• Comparative analysis indicated that the PZM-DSSAE model outperformed five existing state-of-the-art approaches in SPT detection.

Conclusions:

• The developed PZM-DSSAE artificial intelligence model offers a highly accurate and effective tool for the detection of secondary pulmonary tuberculosis.
• This AI-driven approach shows significant potential to improve SPT diagnosis, surpassing current methodologies.
• Further research and clinical validation are warranted to integrate this advanced diagnostic tool into routine practice.