基于深度学习的乳腺癌多级诊断系统的设计与实现

Design and Implementation of Breast Cancer Multi Level Diagnosis System Based on Deep Learning

摘要: 乳腺癌是全球女性发病率最高的恶性肿瘤，传统筛查方法存在敏感性不足、依赖专业医师等问题。特别是对于偏远地区的女性而言，由于当地医疗资源条件、经济情况等因素的影响，从而对乳腺癌的预防和治疗存在一定的延后，很多患者因未能及时地治疗导致病情的恶化。本研究提出了一种基于深度学习的乳腺癌多级诊断系统，并通过对比ResNet50、VGGNeT、GoogleNet和CNN-ViT模型，最终选择准确率为93.47%，精确率为91.35%，召回率为91.50%，F1分数为91.56%的ResNet50模型作为本系统的诊断模型。将模型诊断结果与医生的实际经验相结合，最终得到一份相对准确的诊断报告。患者可在微信小程序及时地知道诊断结果，为下一步治疗提供便利性和快捷性。本系统为乳腺癌的预防和治疗提供了一种可行的方法。

Abstract: Breast cancer is the most prevalent malignant tumor among women globally. Traditional screening methods suffer from insufficient sensitivity and reliance on specialized physicians. Particularly for women in remote areas, factors such as limited medical resources, economic constraints, and geographical access delays prevention and treatment, leading to delayed interventions and disease progression for many patients. This study proposes a deep learning-based multi-stage breast cancer diagnosis system. By comparing ResNet50, VGGNeT, GoogleNet, and CNN-ViT models, we selected ResNet50 as the optimal model, achieving 93.47% accuracy, 91.35% precision, 91.50% recall, and an F1 score of 91.56%. The system integrates model predictions with clinical expertise to generate comprehensive diagnostic reports. Patients can receive real-time results via a WeChat Mini Program, facilitating prompt decision-making for subsequent treatment. This framework offers a feasible solution to improve breast cancer prevention and management, particularly in resource-limited settings.