Pancreatic Cancer Detection Using Quaternion Wavelet Transform and Squeeze-and-Excitation Network with SVM Classifier

Pancreatic cancer (PC) remains one of the most lethal malignancies worldwide, primarily due to the difficulty of early diagnosis and the subtle radiological signatures it presents. To address this challenge, we propose a hybrid computer-aided diagnostic framework that integrates the Quaternion Wavelet Transform (QWT) for robust multi-scale and phase-preserving feature extraction, a Squeeze-and-Excitation (SE) network for adaptive channel-wise feature recalibration, and a Support Vector Machine (SVM) classifier for reliable categorization of pancreatic lesions. The QWT effectively captures discriminative structural information, while the SE network enhances representational quality by modeling inter-channel dependencies. The fused features are subsequently classified by the SVM to ensure efficient and accurate decision-making. Experiments conducted on the publicly available Kaggle CT dataset demonstrate that the proposed method achieves an accuracy of 96.40%, a precision of 95.50%, a recall of 97.05%, a specificity of 96.72%, and an F1-score of 95.73%, outperforming several state-of-the-art approaches. These results highlight the potential of combining QWT, SE networks, and SVM in advancing computer-aided diagnosis for PC and suggest a promising direction for clinical translation.