Thyroid nodule data analysis
Organizer
Speaker
(Tsinghua) Tao Zhou
Time
Monday, November 25, 2024 9:00 PM - 9:30 PM
Venue
Online
Abstract
We obtained data on thyroid ultrasound, pathology and functional indices of selected cases in thyroid surgery from the Third Hospital of North Medicine. All of these cases had one or more thyroid nodules, and our question of interest was how to select from these available data the variables most relevant to the classification of the course of thyroid nodules. We noticed an important fact that a part of the indications of more clinical interest occur infrequently by themselves, and for this reason we have to set very low support thresholds when mining frequent itemsets during the first stage of association rule mining, but this leads to high algorithmic runtimes. Therefore, we have adjusted the computational logic of the association rule algorithm, so that we can mine the frequent itemsets with ultra-low thresholds in a relatively shorter time, and then we can mine the rules that "lead to malignant thyroid cancer" on the basis of these itemsets.