لخّصلي

1.Diagnostic Decision and Retrieval: The framework included a classification stage for diagnostic decisions and a retrieval phase to retrieve relevant cases and their clinical observations from the previous database, aiding radiologists in validating the CAD decision.Multilevel Similarity Measure (MLSM) Algorithm: They introduced an MLSM algorithm based on multiscale information fusion to retrieve relevant cases from a previous patients' database by computing a weighted structural similarity score of multilevel features.Ensemble-SDCNN Model: The ensemble-SDCNN model processed CXR images in parallel to extract both low- and high-level features, enhancing overall classification performance.2.3.4.5.6.

1. 
   

   Proposed Fusion-based Deep Classification Network: They developed a fusion-based deep learning model called ensemble-SDCNN, combining a shallow CNN (SCNN) for low-level features like edge information and a deep CNN (DCNN) for high-level features like TB patterns.

   
   


2. 
   

   Multilevel Similarity Measure (MLSM) Algorithm: They introduced an MLSM algorithm based on multiscale information fusion to retrieve relevant cases from a previous patients’ database by computing a weighted structural similarity score of multilevel features.

   
   


3. 
   

   Data Set and Evaluation: They evaluated their framework’s performance using two well-known CXR datasets and compared it against state-of-the-art methods using metrics like F1 score, average precision, average recall, accuracy, and area under the curve (AUC).

   
   


4. 
   

   Cross-Data Analysis: They conducted cross-data analysis to assess the model's generalizability by training and testing it on different datasets.

   
   


5. 
   

   Diagnostic Decision and Retrieval: The framework included a classification stage for diagnostic decisions and a retrieval phase to retrieve relevant cases and their clinical observations from the previous database, aiding radiologists in validating the CAD decision.

   
   


6. 
   

   Ensemble-SDCNN Model: The ensemble-SDCNN model processed CXR images in parallel to extract both low- and high-level features, enhancing overall classification performance.