Enhancing Diagnostic Accuracy with SE-Inception Model Integration in Pressure Ulcer Detection

Abstract

AIM: Pressure ulcers are a prevalent health concern, often leading to severe complications if not diagnosed and treated promptly. This study introduces the Squeeze-and-Excitation (SE)-Inception model, which integrates SE blocks into the Inception architecture, aiming to enhance classification performance in medical image analysis.    METHODS: The performance of the SE-Inception model was compared to the Xception and Inception v4 models. Key performance metrics such as accuracy, Area Under the Curve (AUC), recall, and Harmonic Mean of Precision and Recall (F1 score) were used to evaluate its efficacy. Gradient-weighted Class Activation Mapping (Grad-CAM) heatmaps were utilized to provide interpretable visual evidence consistent with expert annotations.    RESULTS: The SE-Inception model demonstrated superior accuracy (93%) and AUC (94%), with high recall and F1 scores, indicating its efficacy in reducing false negatives and improving diagnostic reliability.    CONCLUSIONS: Despite the promising outcomes, the study acknowledges the limitation of dataset homogeneity and suggests further validation with diverse datasets for enhanced scalability. The findings support the inclusion of the SE-Inception model in clinical settings to improve diagnostic precision and patient care, particularly in nursing practices for effective pressure ulcer management.