Application of machine learning methods to guide patient management by predicting the risk of malignancy of Bethesda III-V thyroid nodules
Author:
D’Andréa Grégoire1ORCID, Gal Jocelyn2, Mandine Loïc2, Dassonville Olivier1, Vandersteen Clair1, Guevara Nicolas1, Castillo Laurent1, Poissonnet Gilles1, Culié Dorian1, Elaldi Roxane1, Sarini Jérôme3, Decotte Anne4, Renaud Claire5, Vergez Sébastien34, Schiappa Renaud2, Chamorey Emmanuel2, Château Yann2, Bozec Alexandre1
Affiliation:
1. Otorhinolaryngology and Head and Neck Surgery Department, Institut Universitaire de la Face et du Cou, GHS Nice University Hospital—Antoine Lacassagne Centre, Côte d’Azur University , Nice 06103 , France 2. Department of Statistics, Centre Antoine Lacassagne , Nice 06103 , France 3. Otorhinolaryngology and Head and Neck Surgery Department, University Cancer Institute of Toulouse-Oncopole , Toulouse 31400 , France 4. Otorhinolaryngology and Head and Neck Surgery Department, Toulouse University Hospital, Hôpital Larrey , Toulouse 31400 , France 5. Thoracic Surgery Department, Toulouse University Hospital, Hôpital Larrey , Toulouse 31400 , France
Abstract
AbstractObjectiveIndeterminate thyroid nodules (ITN) are common and often lead to (sometimes unnecessary) diagnostic surgery. We aimed to evaluate the performance of two machine learning methods (ML), based on routinely available features to predict the risk of malignancy (RM) of ITN.DesignMulti-centric diagnostic retrospective cohort study conducted between 2010 and 2020.MethodsAdult patients who underwent surgery for at least one Bethesda III-V thyroid nodule (TN) with fully available medical records were included. Of the 7917 records reviewed, eligibility criteria were met in 1288 patients with 1335 TN. Patients were divided into training (940 TN) and validation cohort (395 TN). The diagnostic performance of a multivariate logistic regression model (LR) and its nomogram, and a random forest model (RF) in predicting the nature and RM of a TN were evaluated. All available clinical, biological, ultrasound, and cytological data of the patients were collected and used to construct the two algorithms.ResultsThere were 253 (19%), 693 (52%), and 389 (29%) TN classified as Bethesda III, IV, and V, respectively, with an overall RM of 35%. Both cohorts were well-balanced for baseline characteristics. Both models were validated on the validation cohort, with performances in terms of specificity, sensitivity, positive predictive value, negative predictive value, and area under the receiver operating characteristic curve of 90%, 57.3%, 73.4%, 81.4%, 84% (CI95%: 78.5%-89.5%) for the LR model, and 87.6%, 54.7%, 68.1%, 80%, 82.6% (CI95%: 77.4%-87.9%) for the RF model, respectively.ConclusionsOur ML models performed well in predicting the nature of Bethesda III-V TN. In addition, our freely available online nomogram helped to refine the RM, identifying low-risk TN that may benefit from surveillance in up to a third of ITN, and thus may reduce the number of unnecessary surgeries.
Publisher
Oxford University Press (OUP)
Subject
Endocrinology,General Medicine,Endocrinology, Diabetes and Metabolism
Cited by
2 articles.
订阅此论文施引文献
订阅此论文施引文献,注册后可以免费订阅5篇论文的施引文献,订阅后可以查看论文全部施引文献
|
|