Skin disease diagnosis and treatment model based on text classification algorithm
In response to the challenges of small scale and huge labor cost in biomedical feature modeling in current skin disease assisted diagnosis,as well as the inability to accurately describe the time series of patient disease features,a fusion text classification algorithm is used to integrate commonly used text classification models(TextLSTM,TextCNN,and RCNN)to obtain a model based on transfer learning and neural networks(TLNN model).By extracting the medical features of image sensors and quantizing them,the pretreatment reduces the number of foci and eliminates the feature information with large deviations,thus improving the accuracy of decision data.TLNN model achieves an accuracy of 72.36%on ISIC2018 and PH2 datasets,which is higher than those of the other 3 text classification models.The diagnostic accuracy of TLNN model is close to doctor's diagnosis(92%vs 94%),but the effective diagnostic efficiency is significantly higher than doctor's diagnosis(1.17 min/case vs 4.57 min/case),and the overall efficiency is improved by 290%.The results demonstrate that the fusion text classification algorithm model can obtain accurate diagnosis in less time than the traditional manual diagnosis.TLNN model can be applied to disease diagnosis,and assist doctors in medical decision-making,thereby providing patients with high-quality and convenient intelligent diagnosis and treatment services.
skin diseaseauxiliary diagnosisfusion text classification algorithmDempster-Shafer evidence theorymedical characteristics
NETL
NSTL
万方数据
