Abstract:
Cervical cancer is the most common gynecological malignancy.. Early and accurate identification of the stage of cervical cancer patients can greatly improve the cure rate. In this study, serum sample data were collected from patients with cervical cancer, CIN (cervical intraepithelial neoplasia) I, CIN II, CIN III and hysteromyoma using FT-IR (Fourier-transform infrared spectroscopy) technology. PSO-CNN model for early screening of cervical cancer was designed using a particle swarm algorithm to automatically build a CNN structure with variable number of layers and variable layer class parameters. The experimental results showed that PSO-CNN was the best compared with the classical Lenet, AlexNet, VGG16 and GoogLeNet deep learning models, and the accuracy of PSO-CNN in discriminating five types of samples can reach 87.2%. This study showed that FT-IR technology combined with PSO-CNN model had great potential for non-invasive, rapid and accurate identification of patients with cervical cancer, and can provide a reference for intelligent diagnosis of other diseases.
摘要:
宫颈癌是最常见的妇科恶性肿瘤。.早期准确识别宫颈癌患者的分期可以大大提高治愈率。在这项研究中,从宫颈癌患者收集血清样本数据,CIN(宫颈上皮内瘤变)I,CINII,使用FT-IR(傅立叶变换红外光谱)技术的CINIII和子宫肌瘤。采用粒子群算法设计宫颈癌早期筛查的PSO-CNN模型,自动构建可变层数、可变层类参数的CNN结构。实验结果表明,与经典的Lenet相比,PSO-CNN是最好的,AlexNet,VGG16和GoogLeNet深度学习模型,PSO-CNN对5类样本的判别准确率可达87.2%。本研究表明,FT-IR技术结合PSO-CNN模型在非侵入性、快速准确地识别宫颈癌患者,为其他疾病的智能诊断提供参考。