Perancangan Software Dental Panoramic Berbasis Radiografi Digital Menggunakan Algoritma Feature Matching dan Pyramidal Decomposition untuk Menunjang Kemandirian Alat Kesehatan

Abstract

Background: Dental panoramic radiography (Orthopantomogram/OPG) is a fundamental modality in dentistry. However, the high cost of imported OPG devices hinders widespread access in Indonesia. To support the national Medical Device Independence program (Kemandirian Alat Kesehatan), a cost-effective solution utilizing standard sensors with advanced software reconstruction is proposed. This study aims to evaluate the performance of a custom-built stitching software using Feature Matching and Pyramidal Decomposition algorithms to reconstruct panoramic images from limited Field of View (FOV) sensors.

Methods: This experimental study developed a Python-based imaging pipeline utilizing OpenCV. The process included preprocessing with Contrast Limited Adaptive Histogram Equalization (CLAHE), followed by a comparative analysis of feature detectors: Scale-Invariant Feature Transform (SIFT), Oriented FAST and Rotated BRIEF (ORB), and Accelerated-KAZE (AKAZE). To ensure seamless anatomical transition, image fusion was performed using Pyramidal Decomposition (Multi-band Blending). Performance was measured using Structural Similarity Index (SSIM), Mean Squared Error (MSE), Root Mean Square Error (RMSE), Peak Signal-to-Noise Ratio (PSNR), and computation time. The dataset consists of digital panoramic images that were segmented into several partial images with an overlap of 20–30% to simulate the acquisition process of a moving sensor in an OPG system.

Results: Quantitative evaluation demonstrated distinct performance characteristics. SIFT achieved the highest diagnostic quality with an SSIM of 0.99, PSNR of 38.77 dB, and RMSE of 2.94, proving its superiority in preserving trabecular bone details. ORB provided the fastest processing time at 0.20 seconds but with significantly lower image fidelity (SSIM 0.84, PSNR 26.39 dB). AKAZE showed the lowest performance in this dataset (SSIM 0.75).

Conclusions: The integration of SIFT and Pyramidal Decomposition provides a robust software solution for digital panoramic reconstruction, achieving near-perfect structural similarity (0.99). While ORB allows for real-time preview, SIFT is recommended for final diagnostic imaging in low-cost, indigenous dental imaging systems.