The MVIT-MLKA model, with its complex architecture combining CNNs and Transformers, excels in image feature extraction and capturing long-range dependencies. This gives it strong adaptability and robustness in lesion detection and classification tasks. Compared to traditional machine learning methods and other deep learning models, MVIT-MLKA not only performs better in terms of accuracy, sensitivity, and specificity but also helps reduce inter-observer variability, enhancing diagnostic consistency among physicians.

Although the model showed slight fluctuations in performance on external datasets, it still outperforms other models overall and holds significant potential for clinical applications. With further optimization to improve its generalization capabilities, MVIT-MLKA could become a powerful tool for diagnosing benign and malignant lesions, providing more consistent and accurate support in clinical practice.

Accurate differentiation between benign and malignant pancreatic lesions is critical for patient management. This study aimed to develop and validate a novel deep learning network using baseline computed tomography images to predict benign and malignant pancreatic lesions. This retrospective study across three medical centers constituted a training cohort, an internal testing cohort, and an external validation cohorts. A novel hybrid model, Multi-Scale Large Kernel Attention with Mobile Vision Transformer (MVIT-MLKA), integrating CNN and Transformer architectures, was developed to classify pancreatic lesions. We compared the model's performance with traditional machine learning and deep learning methods. Moreover, we evaluated radiologists' diagnostic accuracy with and without the optimal model assistance.The MVIT-MLKA model demonstrated superior performance for predicting pancreatic lesions, outperforming traditional models and standard CNNs and Transformers. Radiologists assisted by the MVIT-MLKA model showed significant improvements in diagnostic performance compared to those without model assistance, with notable increases in both accuracy and sensitivity. Model interpretability was enhanced through Grad-CAM visualization, effectively highlighting key lesion areas.The MVIT-MLKA model effectively differentiates between benign and malignant pancreatic lesions, surpassing traditional methods and enhancing radiologist performance. This suggests that integrating advanced deep learning model into clinical practice has the potential to reduce diagnostic errors and optimize treatment strategies in clinical practices.

Inclusion Criteria:

All patients with malignant pancreatic lesions had confirmed histopathology according to the 8th edition of the American Joint Committee on Cancer TNM staging system [25]; Lesions were classified as benign if they had either histopathologic confirmation or demonstrated benign characteristics with stability over at least one year of follow-up on CT or MRI imaging; (2) Patients underwent preoperative abdominal contrast-enhanced CT scans; (3) No anti-tumor treatment was conducted before the CT scan

Exclusion Criteria:

(1) Patients with significant motion artifacts or other imaging issues; (2) A time gap of one month or more between the CT scan and subsequent surgery; (3) Tumors less than 10 mm in maximum diameter.

all people were selected from three medical centers in chongqing province

We intend to make the IPD available to qualified researchers upon request. This will be subject to ethical approval and adherence to relevant data protection regulations.