Label-free deep UV microscopy in oral cytology: a step towards stain-free diagnostics.

Abstract

Oral cancer remains a significant global health challenge. Early detection is essential for improving prognostic outcomes, yet current diagnostic practices are hindered by the invasive nature of biopsies and the reliance on staining methods. This study presents a low-cost, label-free deep ultraviolet (UV) microscopy system, integrated with artificial intelligence (AI), for analyzing unstained cytology specimens. Leveraging the absorption properties of nuclei under UV light, this technology produces high-resolution molecular images, enabling real-time, automated, and objective analysis of cellular and nuclear morphology. Forty patients with oral lesions-spanning benign, oral potentially malignant disorders (OPMD), and oral squamous cell carcinoma (OSCC)-participated in this study. Cytology nuclei were segmented using a deep learning-based U-Net architecture, and key nuclear features, including intensity, solidity, eccentricity, and axis ratio, were extracted and analyzed. These features demonstrated high sensitivity (>80%) and specificity (>79%) in distinguishing diagnostic groups. Furthermore, unsupervised clustering based on these features effectively classified patient cohorts, underscoring its potential for early diagnosis. The proposed method eliminates the need for staining, reduces processing time, and minimizes environmental impact, making it particularly suited for primary healthcare settings. By integrating advanced imaging with AI, this scalable approach addresses critical gaps in early oral cancer detection, offering significant potential to improve patient outcomes. Validation in larger and more diverse cohorts is required to enhance its clinical utility.