Institute of Fundamental Technological Research

Simple Summary

Thyroid cancer includes several types that differ in how they grow and how they should be treated. Although ultrasound is widely used to examine thyroid nodules, it can be difficult to determine which type of cancer is present using standard imaging alone. In this study, we applied a computer-based method to automatically measure and analyze ultrasound features of thyroid tumors. By using machine learning techniques, we distinguished between three common types of thyroid cancer: papillary, follicular, and medullary. We found that certain features, such as tumor shape, brightness, and internal structure, were helpful in identifying the cancer subtype. This approach could support doctors in making more accurate diagnoses, reduce unnecessary procedures such as biopsies, and guide more personalized treatment decisions.

thyroid cancer, ultrasound imaging, quantitative analysis, machine learning, papillary thyroid carcinoma, follicular thyroid carcinoma, medullary thyroid carcinoma

Ultrasonography is a basic tool used in the evaluation of thyroid nodules, but there is no single feature of this method which predicts malignancy with statistical significance. The aim of the study is to assess the usefulness of contrast enhanced-ultrasound (CEUS) in the differential diagnosis of thyroid nodules. The highest value of the study results from the multiparameter approach to the evaluation of thyroid lesions in the light of new diagnostics methods and assessment of the unique combinations of both B-mode and CEUS features as predictors of thyroid cancers. Moreover, several qualitative contrast features predicting benign lesions were evaluated. The preliminary results indicate that CEUS is a useful tool in assessing the risk of malignancy of thyroid lesions. The combination of the qualitative enhancement parameters and B-mode sonographic features significantly increases the method’s usefulness. Further studies should be performed to introduce CEUS patterns in the diagnostic algorithm of thyroid nodules.

thyroid cancer, cancer screening, clinical trial, contrast-enhanced ultrasound, thyroid lesion

Introduction:
The latest World Health Organization (WHO) classification from 2022 distinguishes the division of low-risk thyroid neoplasms such as non-invasive follicular thyroid neoplasm with papillary-like nuclear features (NIFTP), follicular tumour of uncertain malignant potential (FT-UMP), and well-differentiated tumour of uncertain malignant potential (WDT-UMP). The final diagnosis is made postoperatively according to histopathologic results. The aim of the study was the assessment of ultrasonographic and cytopathological features of borderline lesions to predict low-risk tumours preoperatively and plan the optimal treatment for that group of patients.

Material and methods:
A total of 35 patients (30 women; 5 men), aged 20–81 years with a mean age of 49 years, were enrolled in the study. The study evaluated 35 focal lesions of the thyroid gland, classified as low-risk neoplasms according to the WHO 2022 classification: FT-UMP (n = 21), NIFTP (n = 7), and WDT-UMP (n = 7). Ultrasonographic features of nodules including contrast-enhanced ultrasound (CEUS) and elastography were assessed by 2 specialists, and the risk of malignancy was evaluated according to EU-TIRADS-PL classification.

Results:
Of the 35 focal thyroid lesions, most were categorised as low or intermediate risk of malignancy according to EU-TIRADS-PL, with dominant category 3 [n = 13 (37.2%)] and category 4 [n = 15 (42.8%)]. High-risk category 5 was assessed in 7 lesions (20%). In cytopathology nodules were categorised as follows (Bethesda System TBSRTC 2023): Bethesda II (n = 4), Bethesda III (n = 2), Bethesda IV (n = 25), Bethesda V (n = 3), and Bethesda VI (n = 1). In the CEUS study, contrasting patterns dominated compared to the surrounding parenchyma, such as enhancement equal to the parenchyma (66.6%) or intense (28.5%), heterogeneous (61.9%), centripetal (42.8%), or diffuse (57.1%) with fast (33.3%) or compared to parenchyma contrast wash-in (42.8%) and its fast (33.3%) or comparable to thyroid parenchyma wash-out (52.3%).

Conclusions:
The study indicates that lesions with uncertain malignant potential typically present features suggesting low to intermediate risk of malignancy based on EU-TIRADS-PL classification, with dominant cytopathologic Bethesda IV category. However, 20% of lesions were assessed tas EU-TIRADS-PL category 5. Low-risk tumours, including NIFTP, FT-UMP, and WDT-UMP, require careful observation and monitoring post surgical treatment due to their potential for recurrence and metastasis. The preoperatively prediction of borderline tumour may play an important role in proper treatment and follow-up.

thyroid tumour, ultrasound, thyroid cancer, contrast-enhanced-ultrasound

Thyroid, Ultrasound, CEUS

Accurate preoperative differentiation of thyroid cancer subtypes is essential for optimizing treatment strategies, particularly in the context of rising incidence rates and increasing use of active surveillance. This study investigates quantitative ultrasound biomarkers based on calcification patterns and tumor margins for distinguishing papillary, follicular, and medullary thyroid carcinomas. The analysis included 110 patients with histopathologically confirmed diagnoses. Tumor boundaries were refined using an energy minimization algorithm, and calcifications were classified by size and location. Margin characteristics were quantified using contrast variance. Each feature demonstrated an AUC > 0.75, with combined biomarkers achieving an AUC of 0.92 for follicular carcinoma detection (sensitivity: 0.87, specificity: 0.76). The results highlight the potential of quantitative assessment of calcifications and margins to enhance ultrasound-based classification of thyroid cancer subtypes.

thyroid cancer, ultrasound, calcifications, tumor margins, quantitative imaging, computer-aided diagnosis