Nathalia Rodrigues Gomes

Background: Actinic cheilitis (AC) is a potentially malignant oral disorder linked to lip squamous cell carcinoma (LSCC). Clinical diagnosis is hindered by lesion heterogeneity. While AI tools have shown promise in oral lesion detection and diagnosis, image segmentation remains underexplored. This study developed and externally validated a deep learning segmentation model for digital photographs of lip lesions. Methods: A total of 489 clinical photographs from one institution were used for training/internal testing, and 82 external images from two independent institutions (UFPB and UFMG) served as external validation. EffiSegNet‑B0 used Image Net‑pretrained EfficientNet‑B0 encoder, with stage features aggregated via Ghost Modules. Manual segmentation masks comprised the clinically affected lip region and were used as ground truth. Performance was assessed using accuracy, sensitivity (recall), precision, Dice Similarity Coefficient (DSC) and intersection over union (IoU). Results: The model achieved high accuracy (> 95%) on both internal and external datasets. DSCs were 0.888 (95% CI: 0.871–0.904) for internal test and 0.875 (95% CI: 0.862–0.888) for external validation, while IoU values were 0.802 (95% CI: 0.777–0.828) and 0.782 (95% CI: 0.763–0.801), respectively. No performance decline was observed on external validation data. The model remained robust to lesion variations (ulceration, crusting, and keratinization). Conclusion: EffiSegNet‐B0 demonstrated high accuracy and adaptability in segmenting lip lesions, reinforcing its potential for clinical image analysis applications. The proposed model generalizes well across different institutions and lesion phenotypes, consistently extracting the complete region of interest. This proof‐of‐concept study supports its promise as a tool for automated lip lesion segmentation in heterogeneous clinical settings.

Background: Artificial intelligence is increasingly applied in digital pathology, yet its reliability in routine diagnostic settings, particularly in fragmented biopsy material, remains limited. We evaluated whether a foundation model could reliably classify head and neck lymphomas and whether integration with established pathological features could improve clinical relevance.Methods: In this multicenter study, 610 cases of head and neck lymphomas were collected across ten institutions. Whole-slide images were digitized and divided into training, validation, test, and two external validation sets. A classification pipeline was developed using UNI (a foundation model) for feature extraction. Nuclear segmentation using CellViT++ enabled extraction of morphometric features, including nuclear area, perimeter, eccentricity, and circularity. Two approaches were compared: a morphology-only model using UNI-derived features and a multimodal model integrating morphometric and clinicopathological variables. Performance was assessed across internal and external validation cohorts. Model behavior was explored using SHAP.Findings: The morphology-only model demonstrated limited performance (accuracy 0.66, F1-score 0.65, AUROC 0.67), with overfitting and poor generalisation. The multimodal model improved performance substantially, achieving accuracy 0.88, F1-score 0.88, and AUROC 0.87 internally, with consistent external validation (accuracy 0.84 and 0.79; AUROC 0.84 and 0.80). SHAP analysis showed that nuclear features were the dominant contributors.Interpretation: Morphology alone imposes a clear performance ceiling in head and neck lymphomas. Multimodal integration provides a more reliable and clinically relevant approach, supporting diagnostic triage in routine practice.

Solitary Fibrous Tumor (SFT) is a rare mesenchymal neoplasm, account approximately 3% of mesenchymal tumors in the oral cavity. We present the case of a 65-year-old female patient with a painless swelling in the left upper lip. Clinical examination revealed a firm, smooth-surfaced submucosal lesion without ulceration, non-tender on palpation, and exhibiting a slightly yellowish hue. Computed tomography demonstrated a well-defined, isodense mass measuring 25 mm in diameter, with no evidence of bone invasion. The lesion was surgically excised, and histopathological analysis confirmed the diagnosis of SFT, supported by positive immunohistochemical staining for CD34 and STAT6. The tumor was classified as low risk for metastasis, and the patient demonstrated favorable postoperative recovery. The patient remains under follow-up with the medical team and has been recurrence-free for 07 months. This case underscores the rarity of oral cavity SFT and emphasizes the importance of comprehensive differential diagnosis to guide effective management and reduce recurrence risk.

Objetivo: The aim of this study is to investigate the clinicopathological and immunohistochemical features of a series of lymphomas involving the parotid glands. Methods: All cases diagnosed in one pathology service from January 2008 and December 2018 were retrospectively retrieved, and the formalin-fixed paraffin-embedded tissue blocks were assessed for diagnostic confirmation. Clinical data were obtained from patients’ medical files. Results: We obtained twelve cases of NHL in the parotid glands representing nine MALT lymphomas, two follicular lymphomas (FL), and one diffuse large B-cell lymphoma, not otherwise specified (DLBCL, NOS). There was a predilection for the female sex (10F:2M), and in all cases it was possible to confirm the involvement of the parotid glands by macroscopic evaluation, imaging studies, or histologic assessment. Clinically, most of the lesions presented as asymptomatic swellings in the parotid region, although associated pain was reported in one case of FL, and three patients with MALT lymphoma had sicca symptoms. Conclusion: In conclusion, NHL affecting the parotid glands are usually of mature B cell lineage, usually representing low-grade subtypes, and frequently simulate other benign or malignant conditions.

Familial florid cemento-osseous dysplasia (FFCOD) is distinct from the sporadic variant and may often be confused with familial conditions presenting with lesions resembling cemento-ossifying fibromas. The current review aims to elucidate the FFCOD variant better and discuss distinguishing features with sporadic florid COD. A review of the literature on FFCOD cases using Google Scholar and Pubmed was performed and summarised. A total of 11 articles with 36 patients were included in the current review. The clinical and radiologic presentations and the pertinent differences from the sporadic variant were discussed. The familial form shows advanced sclerosis and extensive distribution at a younger age, together with impacted teeth and bony expansion in the anterior mandible. Furthermore, distinguishing features from the most important differential diagnoses of other hereditary fibro-osseous conditions, including familial gigantiform cementoma (FGC), hyperparathyroidism jaw tumour syndrome (HP-JTS) and gnathodiaphyseal dysplasia (GDD) are discussed.