医学
门(解剖学)
放射科
淋巴结
神秘的
淋巴
解剖(医学)
转移
逻辑回归
回顾性队列研究
颈淋巴结清扫术
癌
癌症
病理
内科学
替代医学
作者
Megan V. Morisada,Arnaud F. Bewley,Kenneth Broadhead,Reza Assadsangabi,Alireza Paydar,Andrew C. Birkeland,Marianne Abouyared,Lihong Qi,Vladimir Ivanović
标识
DOI:10.1177/19714009231224447
摘要
Background For patients with oral cavity squamous cell carcinoma (OCSCC) without evidence of nodal metastasis (cN0) on pre-operative evaluation, there are no clear guidelines who should undergo elective neck dissection (END) versus clinical surveillance. Objective To identify CT imaging characteristics of sub-centimeter lymph nodes that would help predict the likelihood of nodal metastases on pathology. Methods Retrospective review of cN0 OCSCC patients at a tertiary academic medical center was performed. Inclusion criteria included elective neck dissection, pre-operative CT imaging and presence of metastatic disease within lymph nodes. Control group consisted of patients without nodal metastases on pathology. CT features that were evaluated included asymmetric size, disrupted fatty hilum, asymmetric number, presence of cortical nodule, cortical nodule size, and round/oval shape. We evaluated the associations between CT LN features and the presence of metastases using multi-level mixed-effects logistic regression models. Model evaluation was performed using 5-fold cross-validation. The positive predictive value (PPV) and negative predictive value (NPV) were calculated. Results 26 patients in each study and control groups were included. Three-level mixed-effects logistic regression models indicated round/oval shape (OR = 1.39, p = .01), asymmetric number (OR = 7.20, p = .005), and disrupted fatty hilum (OR = 3.31, p = .04) to be independently predictive in a 3-variable model with sensitivity = 38.0%, specificity = 92.0%, and PPV = 93.8%. Conclusions In cN0 OCSCC patients undergoing END, round/oval shape, asymmetric number, and disrupted fatty hilum of lymph nodes on pre-operative CT imaging are novel and highly predictive of occult nodal disease.
科研通智能强力驱动
Strongly Powered by AbleSci AI