Gastrointestinal Endoscopy Classification using Fuzzy Logic and Machine Learning Algorithms

This effort aims to classify lower gastrointestinal pictures using hyperparameter-tuned machine learning classifiers. Lower gastrointestinal images are essential for diagnosing and treating various gastrointestinal diseases. Categorizing these photographs accurately may help identify and plan therapy, improving patient outcomes. SVM, Random Forest, KNN, Logistic Regression, and Gradient Boosting are used classifiers. Each classifier was evaluated using accuracy, classification reports, and confusion matrices. We used Possibilistic C-Means (PCM) for fuzzy classification to increase accuracy. Medical image processing often involves data uncertainty and imprecision, which PCM helps manage. The method gives membership levels to data points for more flexible and accurate categorization. Hyperparameter tweaking was done using GridSearchCV, ensuring classifier performance was optimal. With optimal values for each model, classifier accuracy varied. The Random Forest classifier has the highest accuracy rate of 69.00%, followed by SVM at 68.25%, Gradient Boosting at 68.00%, KNN at 67.38%, and Logistic Regression at 66.38%. Confusion matrices provided a complete assessment of the models' performance. A detailed summary table showed each classifier's ideal parameters and accuracies, revealing their usefulness in classifying lower gastrointestinal images. This study highlights machine learning classifiers' medical image analysis capabilities. It shows the need of fine-tuning hyperparameters and using fuzzy classification methods like PCM for optimum classification results.