Cutting-edge Deep Learning Solutions for Precise Identification and Definition of Skin Cancer Lesions
This study uses three different types of convolutional neural networks (CNNs), Inception, VGG16, and DenseNet, to look into cutting-edge deep learning methods for accurately identifying the suggested models are trained on a large and varied dataset that includes high-resolution dermatoscopic images...
Saved in:
| Published in: | 2024 International Conference on Innovations and Challenges in Emerging Technologies (ICICET) pp. 1 - 9 |
|---|---|
| Main Authors: | , , |
| Format: | Conference Proceeding |
| Language: | English |
| Published: |
IEEE
07-06-2024
|
| Subjects: | |
| Online Access: | Get full text |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | This study uses three different types of convolutional neural networks (CNNs), Inception, VGG16, and DenseNet, to look into cutting-edge deep learning methods for accurately identifying the suggested models are trained on a large and varied dataset that includes high-resolution dermatoscopic images of different types and states of skin cancer. It is known that the Inception model makes good use of computer resources, and it can also capture features very well. Because it has a deep design, VGG16 is great at picking up small patterns and features inside skin tumors, which makes diagnosis more accurate. Using dense connection patterns, DenseNet encourages information flow and gradient transmission, which helps the network pick up on small differences in the features of lesions. The rough cross-validation and tests on separate datasets are used to see how well these models work. The comparison looks at things like sensitivity, specificity, and general accuracy to see how well each model can find and describe skin cancer spots. The paper discuss these deep learning model and found the accuracy for Inception as 92.63, the VGG16 more close to with accuracy with 84.7 and out of this three model DenseNet proven better accuracy with 93.33. The study also looks into how easy it is to understand the models, using focus maps and feature graphics to show how decisions are made. The findings show that deep learning could help improve the diagnosis of skin cancer, with each design showing its own strengths. The results add to the current discussion about using deep learning for medical picture analysis and also show how these models can be used in real-life therapeutic situations. Ultimately, this study shows how to better and more accurately find skin cancer spots, which will allow for earlier treatment and better results for patients. |
|---|---|
| DOI: | 10.1109/ICICET59348.2024.10616331 |