Decision-Aware Vision Mamba with Context-Guided Slot Mixing for Chest X-Ray Screening and Culture-Based Hierarchical...

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Radiology

Background:

• Chest X-rays are crucial for tuberculosis (TB) diagnosis.
• Differentiating active from inactive TB poses a significant radiological challenge.
• Existing methods struggle with overlapping signs.

Purpose of the Study:

• To develop and validate a novel deep learning framework, Vision Mamba CGSM, for improved TB classification on chest X-rays.
• To enhance the distinction between active and inactive TB.
• To assess the model's performance against established architectures.

Main Methods:

• Implementation of Vision Mamba CGSM, combining a State Space Model (SSM) backbone with a Context-Guided Slot Mixing (CGSM) module.
• Utilizing a hierarchical diagnostic scheme (Normal, Pneumonia, Active TB, Inactive TB).
• Validation on independent test sets and external datasets.

Main Results:

• Achieved 92.96% accuracy and 79.55% Youden Index on the independent test set.
• Demonstrated 97.04% specificity in binary classification of Active vs. Inactive TB.
• Outperformed ResNet152 and ViT-B baselines.

Conclusions:

• Vision Mamba CGSM shows high efficacy in classifying TB stages from chest X-rays.
• The model's feature extraction mechanism generalizes well across datasets.
• This deep learning approach offers a promising tool for radiological diagnosis of TB.