Multi-OSCC

Oral squamous cell carcinoma (OSCC) is among the most prevalent and aggressive malignancies in the head and neck region. Despite advances in therapeutic strategies, early and accurate diagnosis remains a challenge, largely due to the subjective nature of traditional histopathological analysis. Recent progress in digital pathology and machine learning offers promising avenues to enhance diagnostic precision, but these methods require robust and well-annotated datasets for development and validation.

In this study, we introduce a novel dataset of histopathology images derived from OSCC patients treated at Sun Yat-sen Memorial Hospital, Sun Yat-sen University between 2015 and 2022. Patients included in the dataset had a confirmed pathological diagnosis, underwent surgical intervention, and were followed up for a minimum of two years. Tissue samples were processed using standard protocols—including formalin fixation, dehydration, paraffin embedding, sectioning, and H&E staining—ensuring the preservation of critical morphological features.

For each patient, two distinct tissue sections representing the tumor core and its boundary were selected based on pathological evaluation of features such as cancer cell morphology, arrangement, and mitotic activity. Images were captured at magnifications of ×200, ×400, and ×1000 (with a ×10 eyepiece lens) using an Olympus optical microscope, yielding six high-resolution images (2592×1944 pixels) per patient. Alongside the image data, patient-level diagnostic and prognostic annotations were extracted from electronic medical records.

This dataset aims to support the development of automated diagnostic tools and enhance our understanding of OSCC pathology. By providing high-quality, diverse histopathological images along with clinical annotations, we seek to foster research that could lead to more objective and efficient approaches in OSCC diagnosis and prognosis.