Improving The Diagnosis of Thyroid Cancer by Machine Learning and Clinical Data

This repository contains the dataset used in the paper "Improving The Diagnosis of Thyroid Cancer by Machine Learning and Clinical Data" published in Scientific Reports. Please check our formal publication for the full details. The dataset contains 1232 nodules from 724 patients. Each row represents one nodule and each column represents one variable that describes the characteristics of the patient or nodule. The meaning of each variable is summarized below.

id: the unique identity of the patient who carries the nodule

age: the age of the patient

FT3: triiodothyronine test result

FT4: thyroxine test result

TSH: thyroid-stimulating hormone test result

TPO: thyroid peroxidase antibody test result

TGAb: thyroglobulin antibodies test result

site: the nodule location, 0: right, 1: left, 2: isthmus

echo_pattern: thyroid echogenicity, 0: even, 1: uneven

multifocality: if multiple nodules exist in one location, 0: no, 1: yes

size: the nodule size in cm

shape: the nodule shape, 0: regular, 1: irregular

margin: the clarity of nodule margin, 0: clear; 1: unclear

calcification: the nodule calcification, 0: absent, 1: present

echo_strength: the nodule echogenicity, 0: none, 1: isoechoic, 2: medium-echogenic, 3: hyperechogenic, 4: hypoechogenic

blood_flow: the nodule blood flow, 0: normal, 1: enriched

composition: the nodule composition, 0: cystic, 1: mixed, 2: solid

multilateral: if nodules occur in more than one location, 0: no, 1: yes

mal: the nodule malignancy, 0: benign, 1: malignant