Classification and regression trees for the evaluation of thyroid cytomorphological characteristics: A study based on liquid based cytology specimens from thyroid fine needle aspirations

Abstract

Background: This study investigates the potential of classification and regression trees (CARTs) for the evaluation of thyroid lesions. Methods:The study was performed on 521, histologically confirmed cytological specimens prepared via liquid based cytology. For each specimen, contextual and cellular morphology features were recorded by experienced cytopathologists, as described in everyday cytological practice andThe Bethesda System (TBS); these features were subsequently used to construct two CART models, viz. CART-C for the prediction of the cytological diagnosis (according to TBS) and CART-H for the prediction of the histological diagnosis (hereby expressed as either benign or malignant).Results: CART-C had no statistically significant performance from the cytologists' evaluations and CART-H had a very good predictive performance for the histological status.Conclusion: CARTs provide a methodological framework capable for data mining and knowledge extraction. They created simple human understandable rules and classification algorithms that may assist cytopathologists towards decisions based on classification steps, each one linked with a specific risk and moreover by applying cytomorphological characteristics in hierarchical order according to their importance. The two CARTs may be a useful tool for the training of nonexperienced cytopathologists; moreover, they may act as ancillary methods to avoid misdiagnoses and assist quality assurance procedures in the everyday practice of the cytopathology laboratory. K E Y W O R D S classification and regression trees, liquid based cytology, machine learning, morphology, thyroid cytopathology 1 | I N TR ODU C TI ON Fine needle aspiration (FNA) was shown to be a simple, cost-effective, well-established technique in the investigation of thyroid nodules and an optimal tool for clinical management of patients. Cytological classification of thyroid lesions using FNA material provides information concerning the risk of malignancy, thus helping the clinicians to provide appropriate management/treatment of the patient's lesions and reduce unnecessary surgical interventions. Several studies confirm the diagnostic accuracy of FNA, with both high sensitivity (80% to 90.8% 1,2 )and specificity (60% to 100% 3-5 ).The Bethesda 2009 System for Reporting Thyroid Cytopathology (TBS 2009) constitutes a well-established classification system for cytopatholgical evaluation of thyroid lesions. 6-9 Its aim is to guide cytopathologists in the classification of thyroid lesions via established criteria and provides the implied risk of malignancy, in order to guide *The first three authors had equal contribution.

Applications of machine and deep learning to thyroid cytology and histopathology: a review

“…Within thyroid cancer, ML and DL have been applied to other imaging modalities: to diagnose cervical lymph node metastasis in CT images (56) and for computer-aided diagnosis and risk stratification of thyroid nodules in ultrasound scans (57)(58)(59). Although thyroid cancer evaluation is mainly concerned with image analysis, models also exist for diagnosis (26-29), risk stratification (57) and prediction of lymph node metastases (58) from DNA-and RNA-sequencing data.…”

Screening of serum protein biomarkers in hemorrhagic cerebral infarction by surface-enhanced laser desorption/ionization time-of-flight mass spectrometry (SELDI-TOF-MS) technology

“…A decision tree classification algorithm was constructed by using Biomarker Patterns 5.0 software (BPS, Ciphergen Biosystems). BPS is an implementation of the classification regression tree (CART) decision tree system, which uses the peak information generated by the training set of a known sample to construct a binary decision tree algorithm (11). The algorithm allocates each sample in the data set to one of two nodes and makes rules according to the specific peak or strength of the allocator.…”

Digital Medicine in Thyroidology: A New Era of Managing Thyroid Disease

“…Therefore, FNA cytology-based diagnoses are subjective and prone to inter-observer variability. To overcome these problems, several mathematical and computational models have been applied for the diagnostic cytology of thyroid nodules, with reasonable accuracy [21222324252627282930]. AI/ML methods were first applied to the analysis of thyroid FNA specimens by Karakitsos et al [21] in 1996.…”