Accuracy and diagnostic performance of the Bethesda system for reporting thyroid cytopathology in a tertiary endocrine surgical referral center in Belgium.

BACKGROUND: The Bethesda System for Reporting Thyroid Cytopathology is a commonly used classification for fine needle aspiration (FNA) cytology of suspicious thyroid nodules. The risk of malignancy (ROM) for each category has recently been analyzed in three international databases. This paper compares the diagnostic performance of the Bethesda classification in a high-volume referral center in Belgium. METHODS: All consecutive thyroid procedures were registered in a prospective database from January 2010 till August 2022. Patient and surgical characteristics, preoperative Bethesda categories, and postoperative pathology results were analyzed. RESULTS: Out of 2219 consecutive thyroid procedures, 1226 patients underwent preoperative FNA. Papillary thyroid cancer was the most prevalent malignancy (N = 119, 70.4%), followed by follicular (N = 17, 10.1%) and medullary thyroid cancer (N = 15, 8.9%). Micropapillary thyroid cancer was incidentally found in 46 (3.8%) patients. Bethesda categories I, II, III, IV, V, and VI, respectively, represented 250 (20.4%; ROM 4.4%), 546 (44.5%; ROM 3.8%), 96 (7.8%; ROM 20.8%), 231 (18.8%; ROM 15.2%), 62 (5.1%; ROM 72.6%), and 41 (3.3%; ROM 90.2%) patients. Overall ROM was 13.8%. An negative predictive value (NPV) of 96.2% was found. Overall specificity was 64.2% with a positive predictive value (PPV) of 31.9%. Diagnostic accuracy was 67.8%. Compared to international databases (CESQIP, EUROCRINE, and UKRETS), ROM in this study appeared lower for Bethesda category IV (15.2 vs. 26.7% and p = 0.612). CONCLUSION: Despite being validated in numerous studies, ROM based on preoperative FNA cytology classified according to the Bethesda classification may vary among surgical centers and countries as this study reveals a higher NPV and lower PPV.

Classification of Benign-Malignant Thyroid Nodules Based on Hyperspectral Technology.

In recent years, the incidence of thyroid cancer has rapidly increased. To address the issue of the inefficient diagnosis of thyroid cancer during surgery, we propose a rapid method for the diagnosis of benign and malignant thyroid nodules based on hyperspectral technology. Firstly, using our self-developed thyroid nodule hyperspectral acquisition system, data for a large number of diverse thyroid nodule samples were obtained, providing a foundation for subsequent diagnosis. Secondly, to better meet clinical practical needs, we address the current situation of medical hyperspectral image classification research being mainly focused on pixel-based region segmentation, by proposing a method for nodule classification as benign or malignant based on thyroid nodule hyperspectral data blocks. Using 3D CNN and VGG16 networks as a basis, we designed a neural network algorithm (V3Dnet) for classification based on three-dimensional hyperspectral data blocks. In the case of a dataset with a block size of 50 × 50 × 196, the classification accuracy for benign and malignant samples reaches 84.63%. We also investigated the impact of data block size on the classification performance and constructed a classification model that includes thyroid nodule sample acquisition, hyperspectral data preprocessing, and an algorithm for thyroid nodule classification as benign and malignant based on hyperspectral data blocks. The proposed model for thyroid nodule classification is expected to be applied in thyroid surgery, thereby improving surgical accuracy and providing strong support for scientific research in related fields.

A hybrid thyroid tumor type classification system using feature fusion, multilayer perceptron and bonobo optimization.

BACKGROUND: Thyroid tumor is considered to be a very rare form of cancer. But recent researches and surveys highlight the fact that it is becoming prevalent these days because of various factors. OBJECTIVES: This paper proposes a novel hybrid classification system that is able to identify and classify the above said four different types of thyroid tumors using high end artificial intelligence techniques. The input data set is obtained from Digital Database of Thyroid Ultrasound Images through Kaggle repository and augmented for achieving a better classification performance using data warping mechanisms like flipping, rotation, cropping, scaling, and shifting. METHODS: The input data after augmentation goes through preprocessing with the help of bilateral filter and is contrast enhanced using dynamic histogram equalization. The ultrasound images are then segmented using SegNet algorithm of convolutional neural network. The features needed for thyroid tumor classification are obtained from two different algorithms called CapsuleNet and EfficientNetB2 and both the features are fused together. This process of feature fusion is carried out to heighten the accuracy of classification. RESULTS: A Multilayer Perceptron Classifier is used for classification and Bonobo optimizer is employed for optimizing the results produced. The classification performance of the proposed model is weighted using metrics like accuracy, sensitivity, specificity, F1-score, and Matthew's correlation coefficient. CONCLUSION: It can be observed from the results that the proposed multilayer perceptron based thyroid tumor type classification system works in an efficient manner than the existing classifiers like CANFES, Spatial Fuzzy C means, Deep Belief Networks, Thynet and Generative adversarial network and Long Short-Term memory.

[Interpretation of the pathological diagnostic criteria and characteristics of high-grade thyroid follicular-derived carcinoma in the 5th edition WHO classification].

The 5th edition WHO classification of thyroid tumors proposed high-grade non-anaplastic thyroid carcinoma, which includes traditional poorly differentiated thyroid carcinoma (PDTC) and differentiated high-grade thyroid carcinoma (DHGTC), with a prognosis between highly differentiated thyroid carcinoma and anaplastic thyroid carcinoma (ATC), in which about 50% of patients do not take radioactive iodine. Therefore, this classification is of great clinical significance. This article interprets the diagnostic criteria and genetic features of high-grade non-anaplastic thyroid carcinoma in 5th edition WHO classification, comparing with ATC.

Changes in the diagnosis of thyroid tumours in the 5th edition of the WHO classification of endocrine neoplasms.

The WHO classification of thyroid tumours enters its second half-century of development with the 5th edition. Compared to the previous 4th edition of the clas- sification, the permanent increase in information is mainly at the molecular biological level. This has changed the view of very traditional entities - the preferred name for polynodous goiter is (given the monoclonal nature of some nodules) follicular nodular thyroid disease. Some terminological relics have also been re- moved - Hürthle cells are definitively referred to as oncocytes. Follicular adenoma has a new subtype with papillary arrangement (and missing nuclear features of papillary carcinoma). In the already used NIFTP unit, subtypes smaller than 10 mm and oncocytic are newly defined. All oncocytic tumours have an arbitrarily set minimum proportion of oncocytes at 75 %. A multidisciplinary approach to the treatment of thyropathies and the stratification of therapeutic procedures according to risk brought about the introduction of grading into several nosological units of papillary, follicular, and medullary carcinomas. Grading using the number of mitoses determines their quantification at 2 mm² instead of the previously used non-uniform HPFs (high power fields of view). Clarification was made on the basis of genetic findings in a number of other, less frequent diagnoses (e.g. classification of squamous cell carcinoma among anaplastic). Among rare tumors a new category of salivary gland - type carcinomas is formulated with two representatives: mucoepidermoid and secretory carcinoma. Cribriform morular carcinoma previously classified as a variant of papillary carcinoma is newly separated on the basis of the immunological and genetic profile into the newly created category of tumors of uncertain histogenesis. This category also includes sclerosing mucoepidermoid carcinoma with eosinophilia. Microcarcino- ma as a separate entity is not included in the 5th edition. A tumor smaller than 10 mm must be characterized by the appropriate features of the corresponding category. Thyroblastoma replaces terminologically malignant teratoma from the previous classification. Part of the newly established diagnostic criteria is also applicable in FNAB diagnosis. The newly introduced grading in some nosological units can exceptionally change the diagnosis (NIFTP/EFVPTC/non-invasive HG FVPTC), but above all it will affect the choice of therapeutic procedures.

Changes in thyroid cytology reporting in the 3rd edition of the Bethesda system.

Reporting fine-needle aspiration of thyroid nodules in the Bethesda classification is a practice widely used internationally and by us. The revised third edition of the Bethesda System of Reporting Thyroid Cytopathology brings changes in terminology, content, and new chapters. In terms of terminology, an obvious change is the removal of the two-word names of three categories while maintaining the six diagnostic categories of the previous versions - new: BI - non-diag- nostic, BIII - atypia of undetermined significance, BIV - follicular neoplasia. In the detailed description of the findings within the individual categories, the ter- minological changes adopted by the fifth edition of the WHO classification of thyroid neoplasia are respected - in particular, the recommended name follicular thyroid nodular disease for the most frequently represented category BII - benign. In the evaluation itself, the diagnostic specifications accepted by the current WHO classification of histopathological findings are reflected in the individual categories - if they are applicable at the cytological level. Targeted attention will need to be paid to high grade features. The revised version brings new chapters dedicated to molecular testing and evaluation of the paediatric population.

Pregnancy-associated plasma protein A mRNA expression as a marker for differentiated thyroid cancer: results from a "surgical" and a "cytological" series.

PURPOSE: Pregnancy-associated plasma protein A (PAPPA) is a metalloproteinase initially described for its role during pregnancy. PAPPA regulates IGF ligands 1 (IGF1) bioavailability through the degradation of IGF-binding protein 4 (IGFBP4). After the cleavage of IGFBP4, free IGF1 is able to bind IGF1 receptors (IGF1R) triggering the downstream signaling. Recently, PAPPA expression has been linked with development of several cancers. No data have been published on thyroid cancer, yet. METHODS: We evaluated PAPPA, insulin-like growth factor (IGF1), IGF1 receptors (IGF1R) and IGF-binding protein 4 (IGFBP4) mRNA expression levels in a "Surgical series" of 94 thyroid nodules (64 cancers, 16 follicular adenomas and 14 hyperplastic nodules) and in a "Cytological series" of 80 nodules from 74 patients underwent to fine-needle aspiration cytology (FNAC). In tissues, PAPPA was also evaluated by western blot. RESULTS: We found that PAPPA expression was increased in thyroid cancer specimen at mRNA and protein levels and that, adenomas and hyperplastic nodules had an expression similar to normal tissues. When applied on thyroid cytologies, PAPPA expression was able to discriminate benign from malignant nodules contributing to pre-surgical classification of the nodules. We calculated a cut-off with a good specificity (91%) which reached 100% when combined with molecular biology. CONCLUSION: These results show that PAPPA could represent a promising diagnostic marker for differentiated thyroid cancer.

Cancer prevalence in the subcategories of the indeterminate class III (AUS/FLUS) of the Bethesda system for thyroid cytology: a meta-analysis.

PURPOSE: The indeterminate cytologic report represents a major challenge in the field of thyroid nodule. The indeterminate class III of the Bethesda classification system (i.e., AUS/FLUS) includes a heterogeneous group of subcategories characterized by doubtful nuclear and/or architectural atypia. The study aim was to  conduct a systematic review and meta-analysis to evaluate the rate of malignancy in each subcategory of Bethesda III. METHODS: PubMed, CENTRAL, and Scopus databases were searched until April 2020. Original articles reporting data on the subcategories of Bethesda III were included. The histological diagnosis was the reference standard to classify true/false negative and true/false positive cases. RESULTS: The pooled cancer prevalence in each subcategory of Bethesda III was estimated using a random-effects model. Twenty-three papers with 4241 nodules were included. Overall, 1163 (27.4%) were malignant. The cancer rate observed in the subcategories ranged from 15%, in "Hürthle cell aspirates with low risk pattern", to 44%, in "Focal cytologic atypia". CONCLUSIONS: The overall cancer rate found in the Bethesda III ranged more largely than that originally estimated (10-30%) and varied among any scenarios. These evidence-based data represent a reference for the clinical management of these patients.

Outcome and diagnostic reproducibility of the thyroid cytology "indeterminate categories" SIAPEC/SIE 2014 in a consecutive series of 302 cases.

PURPOSE: The clinical impact of the SIAPEC/SIE 2014 classification for thyroid cytology has been addressed in few studies that evaluated the malignancy rate and the relative prevalence of each category. No study analyzed its intra-observer and inter-observer reproducibility, so far. METHODS: We retrospectively collected all "indeterminate" lesions diagnosed before (2011-2014) and after (2015-2018) the application of the SIAPEC/SIE 2014 classification at our Institution. Their relative malignancy risks were calculated based on available histological diagnoses. Cytological and clinical features of TIR3A were compared with the surgical outcome. Finally, a large set of samples was re-evaluated in blind of the original cytological and histological diagnoses by two pathologists, independently. RESULTS: The prevalence of "indeterminate" diagnoses increased in years 2015-2018 (302/1482, 21% with 14% of TIR3A and 7% TIR3B categories) compared to years 2011-2014 (261/1680, 16%). Surgery was performed in 27% TIR3A and in 97% TIR3B cases. Malignancy rates were 40% for TIR3B and 17% for TIR3A, but were greatly influenced by the adoption of the WHO 2017 re-classification of encapsulated follicular-patterned lesions (decreasing to 28% and 6%, respectively). No criteria except for tumor size were associated to malignancy in TIR3A category. Intra-observer agreement of the experienced pathologist was 122/141 (86%), whereas inter-observer agreement between the expert and in-training pathologist was 95/141 (67%). CONCLUSIONS: In this real-life experience, the sub-classification of TIR3A and TIR3B slightly increased the overall prevalence of "indeterminate" diagnoses. Malignancy rates were higher than estimated for both TIR3A and TIR3B categories. Agreement among observers highly depended on pathologist's training.

Thyroid FNA terminology: The case for a single unified international system for thyroid FNA reporting.

The use of reporting terminologies for thyroid FNA cytology enables standardisation and international alignment of the reporting of thyroid cytology results, which is essential. There are currently three major internationally recognised systems: Bethesda (TBS), UK RCPath (Thy), and Italian (TIR). A fourth terminology system used in Japan has identical categories to TBS but with different nomenclature. The aim of this review is to discuss the strengths and weaknesses of the TBS, UK RCPath, and TIR systems, and to make the case for international terminology harmonisation and standardisation.

Thyroid Carcinoma: Phenotypic Features, Underlying Biology and Potential Relevance for Targeting Therapy.

Thyroid carcinoma consists a group of phenotypically heterogeneous cancers. Recent advances in biological technologies have been advancing the delineation of genetic, epigenetic, and non-genetic factors that contribute to the heterogeneities of these cancers. In this review article, we discuss new findings that are greatly improving the understanding of thyroid cancer biology and facilitating the identification of novel targets for therapeutic intervention. We review the phenotypic features of different subtypes of thyroid cancers and their underlying biology. We discuss recent discoveries in thyroid cancer heterogeneities and the critical mechanisms contributing to the heterogeneity with emphases on genetic and epigenetic factors, cancer stemness traits, and tumor microenvironments. We also discuss the potential relevance of the intratumor heterogeneity in understanding therapeutic resistance and how new findings in tumor biology can facilitate designing novel targeting therapies for thyroid cancer.

Presurgical Stratification of Thyroid Nodules - Is it really Needed? Current Guidelines versus Real Life.

Background: Thyroid surgery has various benign and malignant indications. A complete pre-surgical evaluation guides the selection of cases and determines the appropriate extent of the intervention. Minimizing the number of unneeded thyroidectomies could reduce hospitalization costs, as well as post-surgery complications and iatrogenic hypothyroidism. The aim of this study was to retrospectively evaluate the presenting reasons of patients admitted to the hospital for thyroid surgeries and to estimate the need of total thyroidectomies. Methodology: The study included patients admitted in all three Surgical Departments in Timisoara Emergency County Hospital, between January 1st 2018 and December 31st 2019 (2 years). Results: A number of 1036 thyroid surgeries had been performed in 1027 patients and were retrospectively analyzed, comparing the pre-surgical diagnosis with the pathology report. Suspicion of malignancy, compression or functional autonomy was described in 326 /824 total thyroidectomy cases. Cancer was detected in 338 out of the 1027 patients (32.92%), including 39 borderline lesions. A proportion of 80.7% were papillary thyroid cancers. The current presurgical evaluation generated a number of 475 cases displaying differences between the presurgical and postsurgical diagnostic. The phenomenon was observed both in total thyroidectomy and in lobectomy interventions: 22.8% of the lobectomies were diagnosed with thyroid cancer. Conclusion: Our findings confirm that a reliable multidisciplinary approach with standardized presurgical clinical, biochemical and ultrasound evaluation is crucial in patients with indication for thyroid surgery, in order to avoid unnecessary surgeries.

Impact of Noninvasive Follicular Thyroid Neoplasm With Papillary-Like Nuclear Features on Revised Bethesda System Malignancy Rates at a Single Institution.

BACKGROUND: The Bethesda System for Reporting Thyroid Cytopathology (BSRTC) standardizes thyroid cytopathology reporting in six tier diagnostic categories. In recent years, noninvasive encapsulated follicular variant of papillary thyroid carcinoma was reclassified as noninvasive follicular thyroid neoplasm with papillary-like nuclear features (NIFTP). This study examines the impact of NIFTP on the BSRTC risk of malignancy (ROM). METHODS: This was a retrospective review of prospectively collected data from 565 patients who underwent fine needle aspiration and thyroidectomy at a single institution. ROM for each Bethesda category was analyzed and calculated with NIFTP classified as a malignant and nonmalignant lesion. Absolute and relative differences between ROM were compared. RESULTS: Of 565 patients, 19 were Bethesda I, 159 were Bethesda II, 178 were Bethesda III, 46 were Bethesda IV, 42 were Bethesda V, and 121 were Bethesda VI. ROM differences with NIFTP classified as malignant versus nonmalignant for each class were as follows: Bethesda I, no change; Bethesda II, 18%-14%; Bethesda III, 55%-48%; Bethesda IV, 50%-35%; Bethesda V, 93%-91%; and Bethesda VI, 99%-98%. Absolute ROM differences for each category were as follows: Bethesda I, 0%; Bethesda II, 4%; Bethesda III, 7%; Bethesda IV, 15%; Bethesda V, 2%; and Bethesda VI, 1%. CONCLUSIONS: A decreasing trend in absolute and relative ROM was seen in Bethesda II, III, and IV categories; however, exclusion of NIFTP as a malignant lesion did not significantly alter the ROM of BSRTC categories. Surgeons should assess their respective institution's experiences with NIFTP and the BSRTC.

The role of long non-coding RNAs in the pathogenesis of thyroid cancer.

Long noncoding RNAs (lncRNAs) as prominent regulators of gene expression are involved in different layers of expression regulation. These transcripts participate in carcinogenesis of several human malignancies including thyroid cancer. Availability of high throughput techniques such as RNA sequencing and microarray has facilitated identification of lncRNAs whose dysregulation affect tumorigenesis process. Moreover, assessment of differentially expressed lncRNAs between resistant and sensitive cells has led to recognition of biomarkers for therapeutic response. One elucidated aspect of lncRNAs functions is their role in sponging miRNAs. Several miRNA-lncRNA-mRNA triplets have been recognized till now. Any of these triplets is a putative target of interfering with the evolution of cancer. In the current study, we have summarized recent data in the fields of biology of lncRNAs, their role in thyroid cancer and their potential as biomarker or treatment target.

miRNA expression can classify pediatric thyroid lesions and increases the diagnostic yield of mutation testing.

BACKGROUND: Genetic alterations in multiple cell signaling pathways are involved in the molecular pathogenesis of thyroid cancer. Oncogene mutation testing and gene-expression profiling are routinely used for the preoperative risk management of adult thyroid nodules. In this study, we evaluated the potential value of miRNA biomarkers for the classification of pediatric thyroid lesions. PROCEDURE: Double-blind case-control study with 113 resected pediatric lesions: 66 malignant and 47 benign. Quantitative and qualitative molecular data generated with a 10-miRNA expression panel (ThyraMIR) and a next-generation sequencing oncogene panel (ThyGeNEXT) were compared with clinicopathological parameters. RESULTS: miRNAs were differentially expressed in benign versus malignant tumors with distinct expression patterns in different histopathology categories. The 10-miRNA classifier identified 39 (59%) malignant lesions with 100% specificity. A positive classifier score was associated with lymph node metastasis, extrathyroidal extension and intrathyroidal spread. Genetic alterations associated with increased risk for malignancy were detected in 35 (53%) malignant cases, 20 positive for point mutations in BRAF, HRAS, KRAS, NRAS, PIK3CA, or TERT and 15 positive for gene rearrangements involving ALK, NTRK3, PPARG, or RET. The 10-miRNA classifier correctly identified 11 mutation-negative malignant cases. The performance of the combined molecular test was 70% sensitivity and 96% specificity with an area under the curve of 0.924. CONCLUSIONS: These data suggest that the regulatory miRNA pathways underlying thyroid tumorigenesis are similar in adults and children. miRNA expression can identify malignant lesions with high specificity, augment the diagnostic yield of mutation testing, and improve the molecular classification of pediatric thyroid nodules.

Noninvasive Follicular Thyroid Neoplasm With Papillary-Like Nuclear Features in Children: An Institutional Experience and Literature Review.

BACKGROUND: Papillary thyroid carcinoma (PTC) in children has a distinctive set of clinicopathologic features and molecular signature compared to their adult counterparts. The recent recommendation to reclassify encapsulated follicular variant of papillary thyroid carcinoma (EFVPTC) without invasion as noninvasive follicular thyroid neoplasm with papillary-like nuclear features (NIFTP) is based on evidence derived almost exclusively from studies in adults. Clinicopathologic studies restricted to pediatric NIFTP are limited. METHODS: We retrospectively analyzed all pediatric PTC and NIFTP diagnosed and treated in our institution from 1999 to 2016 (n = 31). RESULTS: Using recently published consensus diagnostic criteria, we identified 3 NIFTP and 2 infiltrative follicular variants of papillary thyroid carcinoma (FVPTC) among 31 cases. Two of the NIFTP cases were initially diagnosed as EFVPTC. All 3 patients with NIFTP had unifocal tumors of lower American Joint Committee on Cancer (AJCC) classification (T2 or lower) and were free of lymph node or distant metastasis. Total (n = 1) or completion (n = 2) thyroidectomy was performed in all cases, and only 1 NIFTP patient received subsequent radioablative therapy. No residual or recurrent disease has been observed during follow-up (15-138 months) in patients with NIFTP. CONCLUSIONS: Our experience with NIFTP in children is similar to outcomes reported in adult studies, suggesting that pediatric NIFTP behave indolently as evidenced by the absence of local recurrence in our cohort.

Predictors of malignancy in high-risk indeterminate (TIR3B) cytopathology thyroid nodules.

PURPOSE: The classification of indeterminate cytopathology at thyroid fine-needle-aspiration (FNA) has been updated to reduce the number of unnecessary surgery; the 2014 Italian classification introduced the low-risk (TIR3A) and high-risk (TIR3B) subcategories. Aim of this study was to identify the ultrasonographic (US), clinical and cytological predictors of malignancy among TIR3B nodules from a single institution. METHODS: A prospective observational study including 1844 patients who underwent thyroid FNA from June 2014 to January 2019. Ultrasonographic, clinical and cytological features were recorded. All TIR3B diagnoses were referred to surgery. According to final histology, patients were divided into thyroid cancer (TC) or benign nodules. Chi-square test, or Fisher exact test when appropriate, were used to compare groups and logistic regression analyses were used to determine independent predictors of malignancy. RESULTS: Of 1844 FNAs, 96 (5.2%) were TIR3B. Histology report was available in 65. Among them, 25 (38.5%) were TC. Predictors of TC were nodule size < 20 mm [Odds Ratio (OR) = 5.88, 95% CI 1.91-18.11, p = 0.002], absence or weak intralesional flow [OR = 0.3, 95% CI 0.09-0.77, p = 0.015], microcalcifications [OR = 6.5, 95% CI 1.90-21.93, p = 0.003] at US; nuclear inclusions [OR = 25.3, 95% CI 1.34-476.07, p = 0.031] and chromatin clearing [OR = 3.7, 95% CI 1.27-10.99, p = 0.017] at cytopathology. Patients aged < 55 years had a significantly higher risk of TC [OR = 9.7, 95% CI 2.79-34.07, p < 0.001]. In multivariate analysis, age < 55 and nodule size < 20 mm resulted as independent risk factors. CONCLUSIONS: Patients < 55 years receiving a diagnosis TIR3B on nodules < 20 mm, with microcalcifications, showing specific nuclear atypia at cytopathology are more likely to have TC. Combining US, cytological and clinical features could help determining which patients with a TIR3B diagnosis should be referred to surgery.

Ensemble Deep Learning Model for Multicenter Classification of Thyroid Nodules on Ultrasound Images.

BACKGROUND Thyroid nodules are extremely common and typically diagnosed with ultrasound whether benign or malignant. Imaging diagnosis assisted by Artificial Intelligence has attracted much attention in recent years. The aim of our study was to build an ensemble deep learning classification model to accurately differentiate benign and malignant thyroid nodules. MATERIAL AND METHODS Based on current advanced methods of image segmentation and classification algorithms, we proposed an ensemble deep learning classification model for thyroid nodules (EDLC-TN) after precise localization. We compared diagnostic performance with four other state-of-the-art deep learning algorithms and three ultrasound radiologists according to ACR TI-RADS criteria. Finally, we demonstrated the general applicability of EDLC-TN for diagnosing thyroid cancer using ultrasound images from multi medical centers. RESULTS The method proposed in this paper has been trained and tested on a thyroid ultrasound image dataset containing 26 541 images and the accuracy of this method could reach 98.51%. EDLC-TN demonstrated the highest value for area under the curve, sensitivity, specificity, and accuracy among five state-of-the-art algorithms. Combining EDLC-TN with models and radiologists could improve diagnostic accuracy. EDLC-TN achieved excellent diagnostic performance when applied to ultrasound images from another independent hospital. CONCLUSIONS Based on ensemble deep learning, the proposed approach in this paper is superior to other similar existing methods of thyroid classification, as well as ultrasound radiologists. Moreover, our network represents a generalized platform that potentially can be applied to medical images from multiple medical centers.

Combined use of galectin-3 and thyroid peroxidase improves the differential diagnosis of thyroid tumors.

The differential diagnosis of well-differentiated tumors of follicular cell origin remains the most problematic task in thyroid pathology. Specific morphologic criteria (capsular and/or vascular invasion, nuclear characteristics) are crucial in the diagnosis of these neoplasms. However, the assessment of malignant features is inconclusive in some cases. Moreover, oncocytic thyroid tumors remain controversial in a respect to their pathobiology, behavior and management. Therefore, the useful diagnostic/prognostic thyroid markers are awaited. The aim of our study was to evaluate the expression of galectin-3 and thyroid peroxidase (TPO) in benign and malignant thyroid tumors of follicular cell origin. A total of 186 archival thyroid samples including 38 non-oncocytic follicular adenomas, 53 oncocytic (Hürthle cell) adenomas, 6 non-oncocytic follicular carcinomas, 23 oncocytic (Hürthle cell) carcinomas, 43 non-oncocytic papillary carcinomas, and 23 oncocytic papillary carcinomas were analyzed for galectin-3 and TPO expression by immunohistochemistry. Both types of papillary carcinomas showed significant upregulation of galectin-3 in comparison with the other tumor types, likewise, significant differences in galectin-3 expression were discovered between non-oncocytic and oncocytic variants of studied tumors excluding follicular carcinoma. Significant lowering of TPO was revealed in oncocytic adenomas and papillary carcinomas. In conclusion, the combined use of galectin-3 and TPO markers could help to improve the differential diagnosis of thyroid tumors. Differences in the galectin-3 and TPO expression between some oncocytic and non-oncocytic tumors support their separation in the latest WHO classification of thyroid tumors.

Ultrasound findings of noninvasive follicular thyroid neoplasm with papillary-like nuclear features compared with those of follicular variant of papillary carcinoma and encapsulated papillary carcinoma: a single-institution study in Japan.

Noninvasive follicular thyroid neoplasm with papillary-like nuclear features (NIFTP) is a new entity adopted by the newest World Health Organization classification. It is differentiated from follicular variant papillary thyroid carcinoma (FVPTC) and regarded as non-malignant disease. Here, we compared the ultrasound findings of NIFTP (n = 40) with those of FVPTC (n = 94) and encapsulated PTC (encap-PTC) (n = 157). The NIFTP group showed benign findings on ultrasound significantly more frequently than the FVPTC group based on the Japan Society of Ultrasonics in Medicine criteria: a regular shape (p < 0.001), well edge definition (p = 0.007), smooth character (p < 0.001), isoechoic and homogeneous internal echoes (p < 0.001), lack of punctate microcalcification (p = 0.027), and a regular marginal hypoechoic zone (p < 0.001). Compared to encap-PTC, NIFTP has a significantly higher incidence of benign findings: isoechoic and homogeneous internal echoes (p < 0.001), lack of punctate microcalcification (p < 0.001), and a regular marginal hypoechoic zone (p = 0.004). Based on the ultrasound classification (USC) system at Kuma Hospital, no cases were classified as malignant (USC ≥3.5), but 55.4% of the FVPTCs and 53.5% of the encap-PTCs were diagnosed as malignancy. However, on cytology, the incidence of NIFTP classified as Bethesda-V or -VI (PTC) was very high at 86.9%. All patients underwent surgical treatment, but none of the NIFTP patients showed postoperative recurrence. Although avoiding surgery might be difficult because of the high incidence of malignant cytology, overtreatment (including extensive surgery) for NIFTP can be avoided by paying close attention to the lack of malignant findings on ultrasound.