基于深度学习的黑色素瘤智能诊断多模型算法  

Multi Model Algorithm for Intelligent Diagnosis of Melanoma Based on Deep Learning

作　　者：常炳国 石华龙 常雨馨 CHANG Bing-guo;SHI Hua-long;CHANG Yu-xin(School of Computer Science and Electronic Engineering,Hunan University,Changsha 410082,China)

机构地区：[1]湖南大学信息科学与工程学院,长沙410082

出　　处：《计算机科学》2022年第S01期22-26,177,共6页Computer Science

基　　金：区块链底层技术及应用湖南省重点实验室;湖南省创新型省份建设专项(2020GK2006,2020GK2007)。

摘　　要：皮肤黑色素瘤是一种早期发现可治愈的疾病。目前诊断黑色素瘤的主要方法是基于皮肤镜的人工目视观察,较易受医师医技水平和经验的影响,诊断准确率为75%~80%,且诊断效率低。对此,文中提出一种融合元数据和图像数据的多模态神经网络算法。元数据是通过感知机学习模型提取的患者基本信息、病灶采集部位、图像分辨率和数量的特征向量;图像数据是通过CNN模型提取的特征向量,把两个特征向量进行融合映射以获得疾病分类结果,用于黑色素肿瘤的早期辅助诊断应用。收集整理了ISIC 2019和ISIC 2020的混合数据集,共58457条样本数据,训练样本和测试样本按照4∶1比例进行划分,分别采用所提多模态算法和卷积神经网络方法进行对比实验研究,结果表明,使用所提算法构造的黑色素肿瘤辅助诊断分类模型能够将AUC值提升1%左右,证明其具有一定的使用价值。Skin melanoma is a kind of disease that can be cured by early detection.The main diagnosis method is based on the manual visual observation of dermatoscope.Affected by the doctor’s medical skill and experience,the diagnostic accuracy is 75%~80%and the diagnostic efficiency is low.In this paper,a multi-modal neural network algorithm based on metadata and image data is proposed.Metadata is the feature vector that extracts the basic information of patients,the location of lesions,the resolution and quantity of images through perceptual machine learning model.The image data is extracted from the feature vectors of CNN model,and the two feature vectors are fused and mapped to obtain the disease classification results,which can be used for early auxiliary diagnosis of melanoma.A total of 58457 samples are collected from ISIC 2019 and ISIC 2020 mixed data sets.The training samples and test samples are divided according to the ratio of 4∶1.The multi-modal algorithm and convolutional neural network method proposed in this paper are used for comparative experimental research.The results show that the AUC value of the melanoma auxiliary diagnosis classification model constructed by this algorithm can be improved by about 1%,which has certain use value.

关 键 词：黑色素瘤 皮肤镜 深度学习 多模态算法 元数据 辅助诊断 

分 类 号：TP242.6[自动化与计算机技术—检测技术与自动化装置]