Background: Several studies have used artificial intelligence (AI) to analyze cytology images, but AI has yet to be adopted in clinical practice. The objective of this study was to demonstrate the accuracy of AI‐based image analysis for thyroid fine‐needle aspiration cytology (FNAC) and to propose its application in clinical practice.Methods: In total, 148,395 microscopic images of FNAC were obtained from 393 thyroid nodules to train and validate the data, and EfficientNetV2‐L was used as the image‐classification model. The 35 nodules that were classified as atypia of undetermined significance (AUS) were predicted using AI training.Results: The precision‐recall area under the curve (PR AUC) was >0.95, except for poorly differentiated thyroid carcinoma (PR AUC = 0.49) and medullary thyroid carcinoma (PR AUC = 0.91). Poorly differentiated thyroid carcinoma had the lowest recall (35.4%) and was difficult to distinguish from papillary thyroid carcinoma, medullary thyroid carcinoma, and follicular thyroid carcinoma. Follicular adenomas and follicular thyroid carcinomas were distinguished from each other by 86.7% and 93.9% recall, respectively. For two‐dimensional mapping of the data using t‐distributed stochastic neighbor embedding, the lymphomas, follicular adenomas, and anaplastic thyroid carcinomas were divided into three, two, and two groups, respectively. Analysis of the AUS nodules showed 94.7% sensitivity, 14.4% specificity, 56.3% positive predictive value, and 66.7% negative predictive value.Conclusions: The authors developed an AI‐based approach to analyze thyroid FNAC cases encountered in routine practice. This analysis could be useful for the clinical management of AUS and follicular neoplasm nodules (e.g., an online AI platform for thyroid cytology consultations).

Background: We aimed to determine the indication of fine‐needle aspiration (FNA) for parathyroid adenoma (PA)‐suspected nodules and the cytological features of PA, and to discuss the ancillary techniques for diagnostic confirmation.Method: Clinical, cytological, and histological examinations of 15 PA patients (4.0% of all PA resected patients) were conducted through FNA on 16 nodules. We also examined the cytological preparations of 10 follicular neoplasms (FNs) and 10 poorly differentiated thyroid carcinomas (PDTCs).Results: FNA was performed to detect PA in nine (56.3%) nodules. The remaining seven (43.8%) nodules underwent FNA for lesions considered as thyroid nodules or lymph nodes. The levels of parathyroid hormone (PTH) in the aspiration needle washout fluid were observably high, except for that from one nodule with unsatisfactory FNA. Cytologically, the incidences of wedge pattern (86.7%) and salt and pepper chromatin (86.7%) in PAs were significantly higher than in FNs and PDTCs. In contrast, the appearance of colloid globules and nuclear grooves was less frequent than that of FNs and PDTCs. GATA‐3 expression was intense in all PAs that immunocytochemistry were performed. Histologically, capsular invasion and/or laceration, tumor seeding, granulation tissue, and fibrosis were observed.Conclusions: When PA localization is unusual or inconclusive despite extensive imaging, FNA may be performed. We asserted that wedge pattern, salt and pepper chromatin, and the absence of colloid globules and nuclear grooves are diagnostic cytological indicators of PA rather than of FN or PDTC. We recommend PTH measurements using needle washout fluid for PA‐suspected nodules, and immunocytochemistry with the GATA‐3 antibody for cytologically PA‐suspected nodules.

Background: Clinical management of follicular neoplasms (FNs) using molecular testing of thyroid‐aspirated materials is not routinely performed in Japan. This article aims to identify low‐risk FN nodules that can be followed up without molecular testing.Methods: The relationship between preoperative findings, factors influencing surgical decision, and the risk of malignancy (ROM) was examined in 356 thyroid nodules with cytological diagnosis of FN at Kuma Hospital from January to December 2020.Results: ROMs of FN with cytology results favouring malignancy (41.2%) were significantly higher than those favouring benign (7.7%) or borderline (8.2%) (p < .001). Moreover, ROMs of FN with ultrasonography results of high suspicion (54.5%) were significantly higher than those with low (4.5%) or intermediate suspicion (0%) (p < .0001). There was a large difference in overall ROM in tumours bordering 30 mm in size (<30 mm; 3.6%, ≥30 mm; 20.0%). ROMs of FNs with a tumour volume doubling rate (TVDR) of 1.0/year or more (28.6%) were higher than those of FNs with a lower TVDR (9.9%) (p < .05). The ROMs of FNs with or without one or more of the following four findings suggestive of malignancy: cytological findings favouring malignancy, ultrasonography findings of high suspicion, tumour size ≥30 mm, and TV‐DR ≥1.0/year, were 14.6% and 1.0%, respectively.Conclusion: FNs with no cytological findings favouring malignancy, no ultrasonography findings of high suspicion, tumour size <30 mm and TV‐DR <1.0/year, are considered low risk and can be followed up without the need for molecular testing.

Parathyroid and thyroid lesions appear morphologically similar in cytological smears, and their differentiation can be difficult. The purpose of this study was to determine the diagnostic value of T-cell-specific transcription factor GATA-3 as a marker of parathyroid differentiation in cytology specimens, and to examine the utility of liquid-based cytology (LBC). Cytology smears obtained from surgically removed parathyroid and thyroid specimens, including 15 normal parathyroid glands, 12 cases of parathyroid hyperplasia, 55 parathyroid adenomas, 2 follicular thyroid adenomas, and 3 papillary thyroid carcinomas, were examined by immunocytochemistry using antibodies against GATA-3, parathyroid hormone (PTH), chromogranin A, and thyroid transcription factor 1 (TTF-1). All normal and hyperplastic parathyroids and 98.2% of parathyroid adenomas were positive for GATA-3, while 33.3%, 66.7%, and 60.0% of them, respectively, were positive for PTH. The positive rates for chromogranin A among normal parathyroids (80.0%) and parathyroid adenomas (87.3%) were lower than those for GATA-3. At the same time, all thyroid-derived tumours were positive for TTF-1 and negative for GATA-3, PTH, and chromogranin A. LBC smears of 35 parathyroid lesions indicated that the positive rates for GATA-3, PTH, and chromogranin A were 97.1 %, 97.1%, and 100%, respectively, while in conventional smears, those for PTH (25.5%) and chromogranin A (78.7%) were significantly lower (p < 0.01). Our results suggest that GATA-3 is a more reliable biomarker than PTH or chromogranin A in differentiating parathyroid from thyroid lesions in cytology smears and that LBC is useful in detecting cytoplasmic antigens such as PTH and chromogranin A.