RT info:eu-repo/semantics/doctoralThesis
T1 Evaluación clínica del uso de la inteligencia artificial, tiroglobulina ultrasensible y terapia supresiva de TSH en el carcinoma diferenciado de tiroides
A1 Fernández Velasco, Pablo
A2 Universidad de Valladolid. Escuela de Doctorado
K1 Cancer
K1 Thyroid Cancer
K1 Cáncer de tiroides
K1 Artificial Inteligence
K1 Inteligencia Artificial
K1 32 Ciencias Médicas
AB Differentiated thyroid cancer (DTC) is the most common endocrine cancer, and itsincidence has increased in recent decades due to the widespread use of imaging tests.Nevertheless, DTC mortality has remained stable. The traditional treatment includestotal thyroidectomy, radioactive iodine ablation, and TSH suppression therapy (stTSH)with levothyroxine. However, current guidelines recommend less aggressive approachesbased on dynamic risk stratification (DRS), adjusting treatment according to diseaseprogression. Thyroglobulin (Tg) was traditionally measured after stimulation withrecombinant TSH (rhTSH). The development of ultrasensitive assays (hsTg) allows forearlier detection of recurrences without the need for stimulation tests. Ultrasound isessential for evaluating thyroid nodules, although it presents high inter-observervariability. Artificial intelligence (AI) has improved diagnostic accuracy, reducingunnecessary biopsies and increasing agreement between observers.The main objective of this work was to evaluate and validate new tools to optimize themanagement of DTC, focusing on three key areas: the use of AI in the diagnosis andclassification of thyroid nodules, the evaluation of hsTg's predictive capacity to identifylong-term recurrences, and the adequacy of stTSH during follow-up through DRS.The first article evaluated the clinical impact of a decision support system based on AI(DSS-AI) in the classification of thyroid nodules according to the ACR TI-RADSsystem. A total of 172 patients with thyroid nodules were included, and the diagnosticperformance of six endocrinologists was analyzed before and after using DSS-AI. Theresults showed a significant improvement in sensitivity, specificity, and diagnosticaccuracy after the use of AI, with an increase in the ROC area under the curve (AUC)from 0.776 to 0.817. Furthermore, AI reduced inter-observer variability and reclassifiedmore than 50% of the nodules into lower-risk categories.The second article evaluated the predictive value of basal hsTg compared to rhTSH-Tgfor predicting long-term response in a cohort of DTC patients. A total of 114 patientswere included, and a strong correlation was observed between hsTg and rhTSH-Tglevels. The results showed that hsTg has greater predictive capacity for an excellentresponse, with an AUC of 0.969 compared to 0.944 for rhTSH-Tg. In patients withhsTg determination, the stimulation test provided no additional relevant information.The third article evaluated the evolution of stTSH in a cohort of 216 DTC patientsfollowed for an average of 6.9 years. At diagnosis, 69.2% of patients were at low risk ofrecurrence, compared to 13.6% with high risk. DRS allowed the classification ofpatients with an excellent response from 60% at the start to 70.7% by the end of follow-up. Factors associated with maintaining stTSH included younger age at diagnosis,higher initial risk of recurrence, multifocality, and vascular invasion.Conclusions: The use of DSS-AI improves diagnostic performance in the evaluation ofthyroid nodular pathology; in addition, AI reduced inter-observer variability andimproved agreement. hsTg is a reliable predictor of long-term response in DTC, whilerhTSH stimulation provides no additional relevant information. Although guidelinessuggest relaxing stTSH in low-risk patients, 30.7% continued on suppressive therapy,decreasing to 16.3% after 6.9 years of follow-up. Factors associated with maintainingtherapy were younger age, higher initial risk of recurrence, and DRS evolution.
YR 2024
FD 2024
LK https://uvadoc.uva.es/handle/10324/75243
UL https://uvadoc.uva.es/handle/10324/75243
LA spa
NO Escuela de Doctorado
DS UVaDOC
RD 04-abr-2025