KCGGC: Keypoint Confidence-Guided Gamma Correction for Automatic Enhancement of Lateral Cervical Spine X-ray Images
| Autor: | ZHANG, M., ZHANG, F. |
|---|---|
| Jazyk: | angličtina |
| Rok vydání: | 2024 |
| Předmět: | |
| Zdroj: | Advances in Electrical and Computer Engineering, Vol 24, Iss 2, Pp 93-100 (2024) |
| Druh dokumentu: | article |
| ISSN: | 1582-7445 1844-7600 |
| DOI: | 10.4316/AECE.2024.02010 |
| Popis: | When clinically reviewing lateral cervical spine X-ray images, manual adjustment of contrast is often necessary to highlight features of interest. Gamma correction is one of the most widely used techniques for medical image enhancement in such scenarios. In emulation of radiologists' manual adjustments, this study presents a medical image enhancement scheme guided by keypoint detection confidence to automate the improvement of imaging quality for specific vertebrae in lateral cervical spine images. This method initially generates an enhancement vector to store enhanced images under different gamma correction levels. A detector for detecting 34 morphological keypoints of the cervical spine was trained on a self-constructed CLX-34 dataset, and the optimal gamma correction parameter was determined based on the maximum weighted average confidence of all keypoints across the enhanced images. The proposed weighted average confidence of keypoints metric allows flexible adjustment to enhance focus on regions of interest. Experimental results confirm that the proposed method can improve the readability of lateral cervical spine X-ray images without manual intervention, particularly overcoming the common issue of poor imaging quality of the C7 vertebra. We provide open access to the CLX-34 dataset and pre-trained tools developed in this study. |
| Databáze: | Directory of Open Access Journals |
| Externí odkaz: |
|