Redefining dental image processing: De-convolutional component with residual prolonged bypass for enhanced teeth segmentation
Dental diseases have risen in the past few years due to improper hygiene. Early detection and diagnosis can control this rapid growth in dental diseases. Therefore, different traditional techniques are employed for the detection of dental problems. However, these classical techniques such a...
Saved in:
| Main Authors: | , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Faculty of Technical Sciences in Cacak
2025-01-01
|
| Series: | Serbian Journal of Electrical Engineering |
| Subjects: | |
| Online Access: | https://doiserbia.nb.rs/img/doi/1451-4869/2025/1451-48692502281P.pdf |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Dental diseases have risen in the past few years due to improper hygiene.
Early detection and diagnosis can control this rapid growth in dental
diseases. Therefore, different traditional techniques are employed for the
detection of dental problems. However, these classical techniques such as
X-Ray and CT scans are considered to be time-consuming, ineffective, and
prone to errors due to human intervention. Hence, AI techniques are used to
obtaining precise outcomes for dental-related issues. The conventional ML
(Machine Learning) techniques are inefficient for obtaining enhanced
outcomes as the efficiency of ML techniques heavily depends on image
processing approaches. They are performed and also the quality of the
features that have been extracted. Further, ML techniques lack in producing
better outcomes while dealing with huge datasets. Therefore, the proposed
model employs DL (Deep Learning) techniques due to its capability to learn
the features strongly from the data by using a general-purpose learning
procedure. So, DL techniques can work efficiently on huge datasets. The
proposed DC (De-convolution Component) with RES (Residual Prolonged Bypass)
is employed in the present research work as it is responsible to increase
the spatial resolution of the feature maps and helps in recovering lost
spatial information during the down sampling process. Likewise, the RES
model aids in proficiently proliferating both low-level and high-level
features to the deep layers, which help in generating better-segmented
images. RES model includes prolonged bypass paths that carry feature
information across multiple layers. This ensures that features extracted at
earlier layers (low-level features) are available at much deeper layers.
Implementation of the present research work contributes to enhancing the
overall performance and effectiveness in detecting and diagnosing various
dental issues and possesses the capability to work on both small and massive
datasets effectively. Also, the proposed work contributes to deliver better
accuracy, IoU (Intersection Over Union) and Dice coefficient, compared to
Multi-Headed CNN and Context Encoder-Net, thereby assisting dental
professionals in the detection and diagnosis of various dental issues due to
the effectiveness of the proposed model. |
|---|---|
| ISSN: | 1451-4869 2217-7183 |