Presenting an Automatic Hierarchical Method to Segmentation of Pulmonary Nodules in CT-Scan Images | ||
| AUT Journal of Electrical Engineering | ||
| دوره 57، Issue 2 (Special Issue)، 2025، صفحه 409-420 اصل مقاله (1.28 M) | ||
| نوع مقاله: Research Article | ||
| شناسه دیجیتال (DOI): 10.22060/eej.2024.23555.5623 | ||
| نویسندگان | ||
| Marzia Asadi؛ Hamid Hassanpour* | ||
| Faculty of Computer Engineering and IT, Shahrood University of Technology, Shahrood, Iran | ||
| چکیده | ||
| Pulmonary nodules are significant factors in the development of lung cancer, making timely identification crucial for effective patient management. This study introduces a fully automated method employing a two-stage deep learning approach for the segmentation of pulmonary nodules in CT scan images. In the first stage, we utilize a deep learning network known as BCD-Unet to extract the lung region from the CT scans. The second stage employs another deep learning network that incorporates attention mechanisms and residual modules for the precise segmentation and extraction of pulmonary nodules within the identified lung region. The proposed algorithm was rigorously tested on the LUNA16 dataset, yielding impressive results: a Dice coefficient of 97.75 for lung segmentation, alongside a Dice coefficient of 91.73% and a sensitivity of 92.31% for nodule segmentation. These findings underscore the effectiveness of the proposed method in enhancing the accuracy of pulmonary nodule detection, which is vital for early intervention in lung cancer cases. Also, this research contributes significantly to the field of medical imaging and has important implications for clinical practice in lung cancer management. | ||
| کلیدواژهها | ||
| Attention-Unet؛ BCD-Unet؛ Histogram Equalization؛ Image Segmentation؛ Lung Nodules | ||
| مراجع | ||
|
| ||
|
آمار تعداد مشاهده مقاله: 631 تعداد دریافت فایل اصل مقاله: 556 |
||
| تعداد نشریات | 9 |
| تعداد شمارهها | 455 |
| تعداد مقالات | 5,773 |
| تعداد مشاهده مقاله | 8,414,152 |
| تعداد دریافت فایل اصل مقاله | 6,974,725 |