Background : Myoepithelioma is a rare benign tumor composed of neoplastic myoepithelial cells, and accounts for approximately 1% of all salivary gland tumors. It arises in the parotid gland or palatal salivary glands, predominantly in the 5^th decade of life. Herein, we report two cases of myoepithelioma.

Cases : Case 1 was a male patient in his 60’s. A tumor in the left subaural region was resected and determined histopathologically as being benign 3 years ago. The patient visited our hospital again when he noticed regrowth of the tumor. FNA of the tumor was performed. Case 2 was a male patient in his 20’s. The patient had noticed a mass in the right submandibular region and visited our hospital for surgery ; FNA was performed. The cytology in Case 1 revealed an agglomeration of comparatively uniform cells containing round to short fusiform nuclei with nuclear vacuoles, based on which we made the diagnosis of benign myoepithelial tumor. Cytology in Case 2 revealed cells, mainly in clusters, with heterogeneous atypical nuclei. Also, nuclear vacuoles were observed, the findings suggestive of a myoepithelial tumor.

Conclusions : We wish to underscore the need for recognizing that myoepitheliomas can show a variety of cellular features, and that differential diagnosis of myoepithelioma from pleomorphic adenoma is complicated ; therefore, cytologic estimation of myoepitheliomas should be carried out with caution.

Background : Giardiasis is an infection caused by the protozoan Giardia lamblia, which parasitizes the gastrointestinal and biliary systems. Herein, we report a case in which Giardia lamblia was detected in an endoscopically collected bile cytology specimen.

Case : The patient was a woman in her 90 s who visited our hospital with the chief complaints of chest tightness, pain, and loss of appetite. Abdominal computed tomography showed abnormal shadows in the bile duct. Duodenal ulcer and gastrointestinal bleeding were also seen. After the ulcer healed, we performed a biopsy at the site of the hilar bile duct stricture and collected specimens of bile. Bile cytology specimens stained with Giemsa and PAS clearly showed the characteristic trophozoites of Giardia lamblia, with multiple flagella and 2 nuclei.

Conclusion : Giardiasis is not frequently reported in Japan. However, it should be noted that the number of infected elderly people in Japan who have never traveled abroad is increasing. In this case reported herein, the parasite was not observed in Papanicolaou-stained specimens, because the sample came off during staining. However, we routinely perform Giemsa staining and PAS staining after air-drying the specimens, which is associated with a lower likelihood of less cell detachment and higher likelihood of retention of the characteristic morphology. In order to avoid overlooking parasitic infections, it is desirable to use not only tissue biopsy but also cytology with well-designed sample processing.

Background : We report a case of small cell neuroendocrine cell carcinoma (SCNEC) of the uterine corpus in which intraoperative cytology contributed to the diagnosis.

Case : Imaging findings suggested uterine cancer ; however, preoperative histological diagnosis could not be made because of cervical stenosis. Intraoperative frozen section diagnosis was malignant spindle-cell tumor, based on the cell shape and presence of mitotic figures, whereas intraoperative ascites and stamp cytology suggested neuroendocrine carcinoma, based on the salt-and-pepper pattern of the nuclear chromatin with cell-to-cell adhesion. On the basis of the histological and immunohistological findings, we made the final diagnosis of SCNEC.

Conclusion : Herein, we describe the cytologic features of SCNEC and its differential diagnosis.

At present, five different reporting systems for fine-needle aspiration (FNA) cytology are available in Japan : The “Papanicolaou (class) classification” ; the “Guidelines of the Papanicolaou Society of Cytopathology” (GPSC) ; “The Guidelines of the Japan Thyroid Association” ; “The Bethesda System for Reporting Thyroid Cytopathology (BSRTC)” ; and “General Rules for the Description of Thyroid Cancer”. In 1997, we decided to employ a Modified GPSC system in Ito Hospital for thyroid FNA specimens, in which the “Uncertain malignant potential” category was subclassified into favor benign, borderline, and favor malignant. Postoperative histological examination in 2006 and 2018 revealed a 11% and 22% risk of malignancy in the favor benign category, 50% and 26% in the borderline category, and 50% and 34% in the favor malignant category, respectively. Our subclassification system may prove probabilistically useful, at least to some extent, for differentiating between benign and malignant follicular lesions in thyroid FNA specimens. The diagnostic accuracy could be enhanced by the addition of different approaches such as molecular and genomic examinations.

Follicular thyroid lesions include adenomatous goiter, follicular neoplasms, and the follicular variant of papillary carcinoma. It is often difficult to differentiate among these lesions. Recently, a new disease entity, non-invasive follicular thyroid neoplasm with papillary-like nuclear features (NIFTP), has been proposed for neoplasms that were previously diagnosed as non-invasive encapsulated follicular variant of papillary carcinoma. The Bethesda system has since changed the diagnostic criteria for follicular neoplasms. This article describes the cytological differential diagnosis among the various subtypes of follicular neoplasms and the challenges in the diagnosis of NIFTP, to improve the understanding and diagnosis of follicular lesions.

Thyroid fine-needle aspiration (FNA) biopsy is a highly accurate preoperative diagnostic method ; however, in some cases, morphological diagnosis is challenging, and in such cases, immunocytochemistry (ICC) could prove useful.

The procedure for ICC is similar to that for immunohistochemistry. However, antigen retrieval needs to be considered. To ensure accurate observations, it is necessary to understand the characteristics of cytological specimens. Particular attention should be paid to nonspecific reactions to granular cytoplasm and presence of normal follicular cells in ICC, as they can lead to misinterpretation.

There are three methods of preparing specimens for ICC panels : cell transfer, liquid-based cytology (LBC), and cell block preparation. The LBC method of specimen preparation is easy and fast. However, if the LBC samples are not prepared at the time of the FNA, the cell transfer method should be used.

The application of ICC to thyroid FNA can be divided into two categories : for identification of thyroid tumors that can be definitively diagnosed by ICC and for differential diagnosis between thyroid and non-thyroid tumors. The former includes medullary carcinomas, hyalinizing trabecular tumors, and cribriform variant of papillary carcinomas, whereas the latter includes intrathyroid thymic carcinomas, intrathyroidal parathyroid lesions, and metastatic carcinomas.

The effective use of ICC is expected to further improve the diagnostic accuracy of thyroid FNA. ICC should be performed proactively, especially if the clinical management depends on its result.

Objective : The purpose of this study is to introduce AI for differential diagnosis in cytology of the thyroid (ADDICT), which we are currently developing, and to describe thyroid cytology using AI.

Study Design : Data from 139695 microscopic images were used as the materials for this research. Pre-trained Efficient Net-B0 was used as the image classification model. The model was trained using our dataset through 5-fold cross-validation, and the average of the prediction probabilities was used as the final prediction probability for the test data.

Results : The highest accuracy was for benign lesions (PR AUC=0.99), with PR AUC of 0.9 or higher for all tumors, except for poorly differentiated carcinoma (PR AUC=0.42) and medullary carcinoma (PR AUC=0.88). The accuracy rates for the diagnosis of follicular adenoma and follicular carcinoma were 81% and 94%, respectively. Follicular adenoma and follicular carcinoma were distinguished from one another using the results obtained by compression of the features using T-distributed Stochastic Neighbor Embedding (t-SNE) ; furthermore, lymphoma was divided into three groups, and anaplastic carcinoma was divided into two groups. Results of Grad-CAM demonstrated that AI focused on the nuclei of tumor cells. When “undetermined significance” nodules were predicted using AI, 92.3% of the predictions were correct in distinguishing benign from malignant nodules.

Conclusion : AI-based image analysis is expected to be an ancillary technique for nodules classified as “undetermined significance” and “follicular neoplasm.” In the future, we plan to develop an online platform for thyroid cytology diagnosis and receive consultations from around the world.

“General Rules for the Description of Thyroid Cancer” have been published eight times for more than 40 years. The General Rules provide the standardized terminologies and definitions, serving as a useful guide for the diagnosis and treatment of thyroid tumors. These Rules are principally edited along the lines of the “WHO Histological Classification of Thyroid Tumours” and “The Bethesda System for Reporting Thyroid Cytopathology”. However, there are some differences between the Japanese General Rules and these internationally spread standards. We believe that the Japanese Rules describe the most useful procedures for diagnosis and treatment of thyroid tumor patients for daily medical practice in Japan.