Video Presentation Portrait
Abstract
Pleomorphic adenoma (PA) is the most common benign tumour of the salivary glands, accounting for up to two-thirds of all salivary gland tumors. It most frequently occurs in the parotid gland and typically presents as a slow-growing, asymptomatic mass, predominantly affecting females between the third and sixth decades of life. Histologically, PA exhibits a mixed composition of epithelial, myoepithelial, and mesenchymal components embedded in a myxoid or chondroid stroma, with eosinophilic cytoplasm in polygonal myoepithelial cells. Although often well-circumscribed and localized in the superficial lobe of the parotid, PA can extend into deeper regions, including the parapharyngeal space, which complicates surgical management due to the proximity of critical neurovascular structures. This study aims to highlight the histopathological features of pleomorphic adenoma and correlate them with intraoperative findings. Advanced imaging techniques, particularly MRI, are crucial for accurate preoperative assessment, as delayed surgical intervention increases the risk of malignant transformation into carcinoma ex-pleomorphic adenoma, a more aggressive entity associated with recurrence and metastasis. In recent years, machine learning and deep learning models applied to MRI and ultrasound imaging have demonstrated significant potential in differentiating pleomorphic adenoma from other benign and malignant parotid tumours, improving diagnostic accuracy and supporting earlier clinical decision-making. Microscopic findings confirmed characteristic features of PA, including epithelial proliferation, myoepithelial components, haemorrhagic zones, eosinophilic stroma, and incomplete tumour capsules. Additionally, recent AI-assisted radiomic and convolutional neural network approaches have been shown to complement classical histopathology by extracting high-dimensional imaging features that enhance tumour subtype classification and prognostic assessment. When integrated with clinical and imaging data, these machine learning techniques offer promising avenues for reducing diagnostic uncertainty, optimising surgical planning—particularly in relation to facial nerve preservation—and contributing to more personalised management strategies in parotid gland pathology.