A novel explainable deep generative model-aided transfer learning CNN for pelvis fracture detection

In this paper, we present a novel method: an explainable deep generative model-guided transfer learning Convolutional Neural Network (CNN) for detecting pelvis fractures. Numerous research efforts are currently focused on detecting pelvic fractures using advanced intelligent processes. Although these methods have shown impressive accuracy, they frequently result in extended processing times. To address this challenge, we propose a model that integrates three key components of Explainable Artificial Intelligence (XAI) to enhance the speed of fracture detection. The first component is an explainable segment that pinpoints the location of fractures. The second component is a deep generative approach that carefully analyzes images to simplify feature complexity. The final component is a convolutional neural network (CNN), which assists in decision-making by introducing four transfer learning models: InceptionV2, Xception, MobileNet, and MobileNetV2. Among these, the MobileNet-based approach demonstrates the highest accuracy in classifying conditions, achieving an accuracy of 93.43% , a precision of 99.47% , a sensitivity of 99.38% , and an F1 score of 99.42% , based on a dataset of 876 images. These results are surprising, as the model consistently performed exceptionally well in detecting pelvic fractures. This finding is significant because the model consistently showcased superior performance in identifying pelvic fractures. Moreover, the overall performance of the proposed XAI is comparable to that of medical professionals and orthopedic specialists in detecting pelvic fractures. The proposed technique may offer a promising solution to the challenges in pelvic fracture identification. It harnesses the strengths of deep generative models and CNNs to deliver accurate and interpretable results. • Design of a hybrid model consisting of three phases. • Integration of Explainable AI, Deep Generative models, and Detection. • A fast, precise diagnosis with less processing time and better fracture detection. • Demonstrates superior and reliable performance to medical professionals. • Offer robust performance over the noises.