EE Seminar: Enhancement of Panoramic Radiographs
הרישום לסמינר יבוצע בתחילת הסמינר באמצעות סריקת הברקוד למודל (יש להיכנס לפני כן למודל, לא באמצעות האפליקציה)
Registration to the seminar is done at the beginning of the seminar by scanning the barcode for the Moodle (Please enter ahead to the Moodle, NOT by application)
Electrical Engineering Systems Seminar
Speaker: Omri Dan
M.Sc. student under the supervision of Prof. Nahum Kiryati & Dr. Arnaldo Mayer
Monday, 19th January 2026, at 15:00
Room 011, Kitot Building, Faculty of Engineering
Enhancement of Panoramic Radiographs
Abstract
Panoramic radiography is a fundamental dental imaging tool, yet the 2D projection of 3D anatomy introduces inherent artifacts, including ghost images, spinal superposition, and pharyngeal air-space radiolucency. These degradations obscure critical structures and increase diagnostic uncertainty. This work presents a unified framework for artifact correction through deep de-shadowing and physics-informed data synthesis.
Our framework introduces a selective de-shadowing pipeline that treats X-ray density overlays as "shadows." By combining an artifact segmentation architecture with a transformer-based restoration network, this approach recovers underlying bony textures, achieving State-of-the-Art results in an open-source mandible segmentation task (Dice: 0.9764). To advance beyond mask-dependent methods, we propose diff-deshadow, which utilizes ResFusion a diffusion-based restoration model, to achieve mask-free enhancement by leveraging generative priors to attenuate high-contrast overlays without explicit annotations.
To address the scarcity of paired medical data, we developed a CT-to-panoramic synthesis pipeline that simulates fan-beam geometry to generate artifact-specific benchmarks. The effectiveness of the proposed enhancement was validated through a multidisciplinary reader study with three independent experts. Results demonstrate statistically significant improvements ($p < 0.0001$) across key clinical metrics: Diagnostic Confidence, Contrast Visibility, Overall Image Quality, and Artifact Obstructiveness. These findings indicate that mask-free automated enhancement can substantially improve the diagnostic interpretability of panoramic radiographs, potentially reducing the clinical dependency on higher-radiation 3D imaging.
-סמינר זה ייחשב כסמינר שמיעה לתלמידי תואר שני ושלישי-
This Seminar Is Considered A Hearing Seminar For Msc/Phd Students-
