NasalPolyp-CT-Sinonasal: A Clinical Dataset for Nasal and Paranasal Polyp Detection and Segmentation

The dataset was primarily collected from Tajuddin Ahmed Medical College Hospital, Bangladesh, using advanced clinical imaging equipment to obtain high-resolution nasal CT scans. The ethical clearance for this dataset was obtained from the Institutional Ethics Committee of Daffodil International University, and official permission for clinical data collection was granted by the Assistant Director of Tajuddin Ahmed Medical College Hospital. All participants provided informed consent before image acquisition, ensuring full compliance with medical research ethics and patient confidentiality.

For this study, one two-dimensional (2D) slice per CT scan was selected; thus, each sample corresponds to a single 2D grayscale image of 400 × 400 pixels representing one patient. Consequently, the total dataset volume equals the total number of patients included in the analysis. The dataset comprises two primary pathological classes, Anterior Polyp (AP) and Epithelial Polyp (EP), with background and skull regions treated as auxiliary classes. In total, the dataset contains 247 images of Anterior Polyps (AP) and 206 images of Epithelial Polyps (EP), amounting to 453 samples.

As the dataset volume was relatively limited, controlled data augmentation techniques were applied to expand the dataset to approximately 500 images per class (AP and EP), enhancing model robustness and variability while preserving diagnostic integrity. The selected 2D slices comprehensively cover multiple nasal regions, including the frontal, parietal, and lateral lobes, to capture anatomical variations necessary for robust classification.