Discretized data pattern in endoscopic gastritis images using Dynamic Window and Pairwise Gini Criterion.

Area of Science:

• Medical Imaging Analysis
• Computational Pathology
• Data Mining

Background:

• Current gastritis diagnosis relies on invasive endoscopic procedures and pathological evaluation.
• Non-invasive computer-aided visualization studies have explored feature extraction from endoscopic gastritis images.
• Data pattern extraction (discretization) is a crucial pre-processing step for classification but is often time-consuming and compromises accuracy.

Purpose of the Study:

• To investigate the extraction of data patterns from endoscopic gastritis images.
• To develop and evaluate a novel discretization algorithm for gastritis image features.
• To address the trade-off between pre-processing time and error rate in data discretization.

Main Methods:

• Extensive literature search to identify existing studies on gastritis image analysis.
• Implementation of a proposed discretization algorithm on extracted features from endoscopic gastritis images.
• Evaluation of the algorithm's impact on discretization time and error rate.

Main Results:

• The proposed discretization algorithm successfully reduced both discretization time and error rate.
• The process generated good generalization of data patterns from extracted endoscopic gastritis features.
• This approach offers a significant improvement over existing data pre-processing methods.

Conclusions:

• Determining discretized data patterns from endoscopic gastritis images can enhance classification accuracy and reduce learning time.
• The developed algorithm provides an efficient and accurate method for pre-processing gastritis image data.
• This computer-aided approach holds promise for improving non-invasive gastritis diagnosis.