Diffusion-Based Hierarchical Multi-Label Object Detection to Analyze Panoramic Dental X-rays

• DOI: 10.48550/arxiv.2303.06500
• Publication Date: 2023-01-01
• Authors: İbrahim Ethem Ham...; Sezgin Er; Enis Simsar; Anjany Sekuboyina; Mustafa Gündoğar; Bernd Stadlinger; Albert Mehl; Bjoern Menze
• Keywords: FOS: Computer and information sciences; Panoramic Dental X-ray; Diffusion Network; Multi-Label Object Detection; Transformers; Computer Vision and Pattern Recognition (cs.CV); Computer Science - Computer Vision and Pattern Recognition; 0202 electrical engineering, electronic engineering, information engineering; 02 engineering and technology; Hierarchical Learning

Due to the necessity for precise treatment planning, use of panoramic X-rays identify different dental diseases has tremendously increased. Although numerous ML models have been developed interpretation X-rays, there not an end-to-end model that can problematic teeth with enumeration and associated diagnoses at same time. To develop such a model, we structure three distinct types annotated data hierarchically following FDI system, first labeled only quadrant, second quadrant-enumeration, third fully quadrant-enumeration-diagnosis. learn from all hierarchies jointly, introduce novel diffusion-based hierarchical multi-label object detection framework by adapting method formulates as denoising diffusion process noisy boxes boxes. Specifically, take advantage data, our utilizes box manipulation technique in network inference previously trained order. We also utilize efficiently partial annotations give needed information about each abnormal tooth planning. Experimental results show significantly outperforms state-of-the-art methods, including RetinaNet, Faster R-CNN, DETR, DiffusionDet analysis demonstrating great potential partially datasets. The code are available at: https://github.com/ibrahimethemhamamci/HierarchicalDet.