Gingivitis is a prevalent oral health condition that, if left untreated, can progress to periodontitis, significantly impacting masticatory function, phonation, and aesthetics due to potential tooth loss. Given the persistent global disparities in healthcare accessibility, the integration of Artificial Intelligence (AI) for remote diagnostic support in dentistry has become critically important. This paper introduces an innovative deep learning framework for automated gingivitis detection from intraoral images. Our methodology emphasizes a robust data preprocessing pipeline, crucial for optimizing feature extraction and enhancing the model's recognition capabilities. This pipeline includes region of interest (ROI) extraction and meticulous annotation to mitigate confounding elements within the images, thereby improving model reliability. The proposed data processing workflow systematically involves image cropping to define two distinct regions of interest (ROIs), data augmentation via techniques such as rotation, blurring, and noise injection, training two parallel YOLOv9 models, one for precise tooth segmentation, another for binary gingivitis detection and calculating the Intersection over Union (IoU) to identify teeth affected by gingivitis. The workflow also includes coloring segmentation masks (bright for diseased teeth, dark for healthy teeth) and transforming mask and bounding box coordinates from the ROI space back to the original image coordinate system. The experimental results demonstrate that this comprehensive method achieves good performance indicators, including precision, recall, mAP@50:95 for tooth segmentation of 94.8, 94.6, 93.1, respectively, and improved accuracy, sensitivity and specificity for gingivitis detection of 86.79, 84.36, 95.34, respectively, compared to other methods. This research validates the efficacy of AI-driven solutions for accurate and efficient remote gingivitis detection, holding significant promise for improving oral healthcare accessibility.