Sompon Apornvirat

Background: Malignant adenomyoepithelioma (M‐AME) is an extremely rare and morphologically diverse breast tumour, posing significant diagnostic and management challenges. We present seven cases of M‐AME with different histological compositions to contribute additional data to the existing literature. Through these cases, we review and discuss potential pitfalls and classification challenges in diagnosing M‐AME. Methods and results: We performed a retrospective clinicopathological review of seven cases of M‐AME from women aged 45–74 years (median 53 years). Each case exhibited distinct architectural patterns and compositions of classic and malignant components: (1) Adenomyoepithelioma (AME) with metaplastic spindle and squamous cell carcinoma; (2) Invasive AME with an aberrant immunohistochemical profile; (3) Invasive AME with an adenomyoepithelial adenosis pattern; (4) Metaplastic spindle cell carcinoma arising from AME with suspicion of M‐AME in situ; (5) Atypical AME with late lung metastasis harbouring HRAS p.G13R and in‐frame PIK3R1 deletion; (6) and (7) M‐AME in situ. Conclusion: M‐AME encompasses heterogeneous patterns that overlap with the infiltrative‐like pattern of classic AME (adenomyoepithelial adenosis form of tubular AME) and metaplastic breast carcinoma. Recognition of the underlying AME architecture, coupled with immunohistochemical and molecular correlation, supports pragmatic classification into in situ or invasive M‐AME or conventional breast carcinoma arising in the context of AME. Together with clinical, radiological, and molecular correlations, this approach can help guide appropriate management.

Aims To objectively evaluate the prevalence of histological border irregularities in fibroadenomas (FAs) and characterise their association with specimen type, histological subtypes, prior procedures and associated pathologies. Methods A retrospective review of 912 FA specimens (549 core needle biopsies (CNBs), 45 vacuum-assisted biopsies (VABs), 242 excisions and 76 mastectomies) from 816 lesions was conducted. An irregular border was defined as an interface between the FA and adjacent parenchyma that was not smooth or circumscribed. Statistical analyses were performed to identify factors associated with irregularities in the FA contours. Results Border irregularity was present in 20.6% (188/912) of specimens. Prevalence was significantly lower in CNB (8.4%) compared with VAB (31.1%), mastectomy (34.2%) and excision (42.1%). By subtype, irregularity was most frequent in myxoid (41.7%), cellular (35.8%) and hyalinised (28.9%) FAs, compared with complex (20.4%), usual (16.2%) and juvenile (13.3%) forms. Calcifications, carcinoma in situ and invasive carcinoma, occurring within FAs, were associated with increased frequency of irregular borders. In multivariable analysis, a history of prior procedure was the strongest independent predictor of irregularity (OR 3.69, pConclusions While FAs are conventionally described as possessing circumscribed contours, our study found irregular borders in approximately one in five specimens. This feature is significantly associated with prior clinical procedures, specific histological subtypes and certain concurrent pathologies. It is essential to recognise that an irregular border in fibroepithelial lesions can be encountered in FAs and is not exclusive to phyllodes tumours.

Malignant transformation of craniopharyngioma is an exceptionally rare and poorly characterized phenomenon. We report a case of a 15‐year‐old girl who developed symptom recurrence 9 years after undergoing surgery and radiotherapy for adamantinomatous craniopharyngioma. Histology revealed malignant areas comprising highly proliferative basaloid cells with abrupt transitions into ghost cells, resembling ghost cell odontogenic carcinoma. We conducted a literature review and pooled case analysis of 45 cases, which showed a median benign‐to‐malignant latency of 10 years and a median overall survival of 9 months. The majority of benign diagnoses were adamantinomatous type. Prior radiotherapy was noted in 27 cases (60%). Histological findings of malignant components were reported as unspecified malignant transformation in 20 cases (44.4%), squamous cell carcinoma (SCC) in 16 cases (35.6%), odontogenic‐type carcinomas in 5 cases (11.1%), and other types in 4 cases (8.9%). Immunohistochemical analyses showed consistent positivity for pan‐cytokeratin, p63, and p40 in all cases performed. The p53 and Ki‐67 also consistently exhibited increased or abnormal expression. Due to the overlapping histological and immunohistochemical features of odontogenic‐type carcinomas and SCC, careful assessment is recommended when diagnosing recurrent CPs.

This study evaluated ChatGPT's ability to generate clinically integrated multiple‐choice questions (MCQs) directly from digestive pathology lecture slides and assessed the educational quality of the output using expert review, psychometric analysis, and student feedback. Lecture slides covering 30 digestive diseases were uploaded to ChatGPT‐4o using a structured prompt to produce 30 MCQs, each with a clinical vignette, five answer choices, and explanations. Three pathologists and a gastrointestinal and hepatobiliary pathologist evaluated question quality. Psychometric analysis was conducted using difficulty index, discrimination index, and distractor quality based on responses from 31 s‐year medical students. Reviewer and student feedback were also analyzed. Of the 30 MCQs, 28 (93.3%) were rated as excellent by all pathologists. Minor issues such as factual inaccuracies or overly leading clues were found in 11 questions (36.7%) and corrected. Most questions showed acceptable psychometric properties: 23 items (76.7%) met difficulty index thresholds, and 28 (93.3%) met discrimination index thresholds. Of 120 distractors, 54 (45%) were functional. Students reported improved clarity, relevance, and clinical integration. ChatGPT can generate accurate, clinically relevant MCQs aligned with lecture material. With expert review, this approach provides a practical and scalable tool for pathology education in resource‐ and time‐limited settings.

Background: Human epidermal growth factor receptor 2 (HER2) is an important predictive and prognostic biomarker in breast cancer. Immunohistochemistry (IHC) is the preferred initial test due to its cost-effectiveness and simplicity. While fluorescence in situ hybridization (FISH) is the traditional gold standard test for HER2 gene amplification, DISH has emerged as an accepted alternative that allows evaluation under a standard light microscope. Objectives: To evaluate agreement between HER2 IHC (2+/3+) and DISH in node-positive primary breast cancers and compare findings with published IHC and FISH data. Design: Retrospective single-center cohort study using nationwide referral specimens. Methods: Cases of pathologically confirmed lymph node metastasized invasive breast carcinoma with HER2 IHC scores of 2+ and 3+ were retrieved. Interpretation of HER2 IHC was performed using the 2023 ASCO/CAP guideline. HER2 DISH was conducted and evaluated by the HER2/CEP17 signal ratio on primary tumors first, and on metastasized lymph nodes in cases of persistent technical failure. Results: Among 1,307 breast cancers, DISH detected HER2 amplification in 933 cases, including 92% (760) of IHC 3+ cases and 36% (173) of IHC 2+ cases. Seven cases with persistent technical failure on primary tumors were resolved when switching to lymph node specimens. Comparison with the meta-analysis data of IHC and FISH showed no significant differences, indicating that DISH is a reliable alternative to FISH. Conclusion: Our study demonstrates a high concordant rate between HER2 IHC and DISH in the IHC 3+ group and a low positive rate in the IHC 2+ group. We found no significant difference in the positive rates of HER2 IHC to DISH when compared with prior data of IHC to FISH, reaffirming the use of HER2 DISH as an effective and more accessible alternative to FISH in HER2 2+/3+ breast cancer.

The liver's unique cellular structure makes it a frequent site for metastatic cancer. In colorectal liver metastasis (CRLM), surgical resection is essential for long‐term survival. Histopathological growth patterns (HGPs) in CRLM, including desmoplastic and nondesmoplastic patterns, provide critical prognostic information. Tumor budding (TB) and poorly differentiated clusters (PDCs), indicators of aggressive cancer behavior, are evaluated using standardized histological scoring systems and are linked to epithelial‐mesenchymal transition. This study explored the correlation between HGPs, TB, and PDCs in CRLM. Archived data from Thammasat University Hospital, including resected CRLM specimens, were analyzed. This study evaluated 51 CRLM resection specimens treated with preoperative systemic therapy, finding most to be nondesmoplastic with low TB and grade 1 PDC. Desmoplastic growth was significantly more prevalent in cases receiving preoperative chemotherapy than those that did not. Higher 3‐year mortality was noted in nondesmoplastic groups and those with higher TB and tumor regression grade (TRG) scores. Significant correlations were observed between HGPs, TB, and PDCs, despite challenges in assessing these parameters due to issues with noncancer cells, extracellular mucin, bile ductular proliferation, and retraction artifacts. This study underscores the prognostic significance of HGPs, TB, PDCs, and TRG scores in CRLM, highlighting the need for precise histopathological evaluation for more accurate prognostic implications.

Objectives: This research aimed to evaluate the effectiveness of ChatGPT in accurately diagnosing hepatobiliary tumors using histopathologic images. Methods: The study compared the diagnostic accuracies of the GPT-4 model, providing the same set of images and 2 different input prompts. The first prompt, the morphologic approach, was designed to mimic pathologists’ approach to analyzing tissue morphology. In contrast, the second prompt functioned without incorporating this morphologic analysis feature. Diagnostic accuracy and consistency were analyzed. Results: A total of 120 photomicrographs, composed of 60 images of each hepatobiliary tumor and nonneoplastic liver tissue, were used. The findings revealed that the morphologic approach significantly enhanced the diagnostic accuracy and consistency of the artificial intelligence (AI). This version was particularly more accurate in identifying hepatocellular carcinoma (mean accuracy: 62.0% vs 27.3%), bile duct adenoma (10.7% vs 3.3%), and cholangiocarcinoma (68.7% vs 16.0%), as well as in distinguishing nonneoplastic liver tissues (77.3% vs 37.5%) (Ps ≤ .01). It also demonstrated higher diagnostic consistency than the other model without a morphologic analysis (κ: 0.46 vs 0.27). Conclusions: This research emphasizes the importance of incorporating pathologists’ diagnostic approaches into AI to enhance accuracy and consistency in medical diagnostics. It mainly showcases the AI’s histopathologic promise when replicating expert diagnostic processes.