Liao Hongfan, Huang Cheng, Liu Chunhua, Zhang Jiao, Tao Fengming, Liu Haotian, Liang Hongwei, Hu Xiaoli, Li Yi, Chen Shanxiong, Li Yongmei
Department of Radiology, The First Affiliated Hospital of Chongqing Medical University, Chongqing, 400016, China.
College of Computer and Information Science, Southwest University, Chongqing, 400715, China.
Radiol Med. 2025 Apr;130(4):508-523. doi: 10.1007/s11547-025-01949-5. Epub 2025 Jan 20.
Accurate differentiation between benign and malignant pancreatic lesions is critical for effective patient management. This study aimed to develop and validate a novel deep learning network using baseline computed tomography (CT) images to predict the classification of pancreatic lesions.
This retrospective study included 864 patients (422 men, 442 women) with confirmed histopathological results across three medical centers, forming a training cohort, internal testing cohort, and external validation cohort. A novel hybrid model, Multi-Scale Large Kernel Attention with Mobile Vision Transformer (MVIT-MLKA), was developed, integrating CNN and Transformer architectures to classify pancreatic lesions. The model's performance was compared with traditional machine learning methods and advanced deep learning models. We also evaluated the diagnostic accuracy of radiologists with and without the assistance of the optimal model. Model performance was assessed through discrimination, calibration, and clinical applicability.
The MVIT-MLKA model demonstrated superior performance in classifying pancreatic lesions, achieving an AUC of 0.974 (95% CI 0.967-0.980) in the training set, 0.935 (95% CI 0.915-0.954) in the internal testing set, and 0.924 (95% CI 0.902-0.945) in the external validation set, outperforming traditional models and other deep learning models (P < 0.05). Radiologists aided by the MVIT-MLKA model showed significant improvements in diagnostic accuracy and sensitivity compared to those without model assistance (P < 0.05). Grad-CAM visualization enhanced model interpretability by effectively highlighting key lesion areas.
The MVIT-MLKA model efficiently differentiates between benign and malignant pancreatic lesions, surpassing traditional methods and significantly improving radiologists' diagnostic performance. The integration of this advanced deep learning model into clinical practice has the potential to reduce diagnostic errors and optimize treatment strategies.
准确区分胰腺良性和恶性病变对于患者的有效管理至关重要。本研究旨在开发并验证一种使用基线计算机断层扫描（CT）图像预测胰腺病变分类的新型深度学习网络。
这项回顾性研究纳入了三个医疗中心的864例患者（422例男性，442例女性），其组织病理学结果已得到证实，组成了一个训练队列、内部测试队列和外部验证队列。开发了一种新型混合模型，即具有移动视觉Transformer的多尺度大核注意力模型（MVIT-MLKA），该模型整合了卷积神经网络（CNN）和Transformer架构来对胰腺病变进行分类。将该模型的性能与传统机器学习方法和先进的深度学习模型进行了比较。我们还评估了在有无最佳模型辅助的情况下放射科医生的诊断准确性。通过区分度、校准度和临床适用性来评估模型性能。
MVIT-MLKA模型在胰腺病变分类方面表现出卓越性能，在训练集中的曲线下面积（AUC）为0.974（95%置信区间0.967 - 0.980），在内部测试集中为0.935（95%置信区间0.915 - 0.954），在外部验证集中为0.924（95%置信区间0.902 - 0.945），优于传统模型和其他深度学习模型（P < 0.05）。与无模型辅助的放射科医生相比，MVIT-MLKA模型辅助下的放射科医生在诊断准确性和敏感性方面有显著提高（P < 0.05）。Grad-CAM可视化通过有效突出关键病变区域增强了模型的可解释性。
MVIT-MLKA模型能有效区分胰腺良性和恶性病变，超越了传统方法，并显著提高了放射科医生的诊断性能。将这种先进的深度学习模型整合到临床实践中有可能减少诊断错误并优化治疗策略。
