基于PET/CT图像融合的非小细胞肺癌淋巴结转移预测

Prediction of lymph node metastasis in non-small cell lung cancer based on PET/CT image fusion

淋巴结转移是决定肿瘤分期和预后的主要因素。因此,准确评估淋巴结转移状况对患者的治疗、预后和生存状况具有重要意义。目前的研究大多需要医生勾画肿瘤进行预测,提供准确勾画肿瘤的图像是耗时且费力的。为解决这一问题,提出了一种基于PET/CT图像融合的非小细胞肺癌淋巴结转移诊断方法,即只需病人拍摄的PET/CT图像,无需医生勾画肿瘤。该方法采用直接阅片的方式用Resnet18从影像提取特征向量经过融合模块将PET信息和CT信息融合,分类得到最终结果。融合模块通过学习来确定CT特征和PET特征在不同区域的重要性,分配不同的权重融合。该模型在NSCLC Radiogenomics数据集上的准确度为0.795,敏感性为0.6。在患者淋巴结转移的分类方面,提高了肺癌患者的诊断准确率。

Lymph node metastasis is a major factor in determining tumor stage and prognosis.Therefore,accurate assessment of lymph node metastasis status is of great significance for the treatment,prognosis,and survival of patients.Most of the current research requires doctors to delineate tumors for prediction,and providing accurate images of tumors is time-consuming and laborious.In order to solve this problem,this paper proposed a diagnostic method for lymph node metastasis of non-small cell lung cancer based on PET/CT image fusion.Only PET/CT images taken by the patient were used,and the doctor did not need to delineate the tumor.In this method,the feature vectors obtained from the image by Resnet18 were obtained by direct reading,and the PET information and CT information were fused through the fusion module,and the final result was obtained by classification.The fusion module determined the importance of CT features and PET features in different regions through learning,and assigned different weights to fusion.The accuracy of the model on the NSCLC Radiogenomics dataset is 0.795 and the sensitivity is 0.6.In terms of the classification of lymph node metastasis in patients,the diagnostic accuracy of lung cancer patients is improved.