Tools
Tools
Sonawane, Raj Shrikant (2023) Skin-Cancer Classification Using Deep Learning with DenseNet and VGG with Streamlit-Framework Implementation. Masters thesis, Dublin, National College of Ireland.
Preview |
PDF (Master of Science)
Download (1MB) | Preview |
Preview |
PDF (Configuration master)
Download (2MB) | Preview |
Abstract
Skin cancer is consistently ranked among the top three most lethal tumors that are caused by DNA damage and can lead to death on a worldwide scale. The growth of the epidermal layer of the skin is abnormal or rapid, which encourages the formation of tumors in the body. After being identified, this has had an effect on people of all ages and has a reputation for being an expensive ailment to treat in the past. An earlier study that looked at the various types of skin cancer found that earlier detection might result in quicker treatment and lower overall costs. This research project aims to build Deep Learning models that will assist in classifying skin lesions as either benign or malignant. The models will be used in the context of a skin cancer diagnosis. The application of deep learning will help cut both the amount of time and money spent on therapy. Using the Streamlit application development framework, a web-based application has been developed with the intention of bridging the gap between Deep Learning and medical expertise. The dataset compiled by the International Skin Image Collaboration and hosted on Kaggle has been utilized for the purpose of training and testing algorithmic models. There was no evidence of class imbalance in the data; however, space limitations necessitated the application of a variety of Data Augmentation strategies. For categorization, deep learning models including DenseNet-121, DenseNet-169, DenseNet-201, VGG- 16, and VGG-19 have been created. By considering and striking a balance between the different metrics that are necessary in the field of medicine, DenseNet-121 with three layers was able to achieve scores of 98.24 and 70.55 for its sensitivity and specificity, respectively, and 84.64 for its accuracy. The findings demonstrated that models’ performance can be improved by adjusting their hyper-parameters.
| Item Type: | Thesis (Masters) |
|---|---|
| Supervisors: | Name Email Cosgrave, Noel UNSPECIFIED |
| Subjects: | Q Science > QA Mathematics > Electronic computers. Computer science T Technology > T Technology (General) > Information Technology > Electronic computers. Computer science R Medicine > RC Internal medicine > RC0254 Neoplasms. Tumors. Oncology (including Cancer) R Medicine > RL Dermatology Q Science > Q Science (General) > Self-organizing systems. Conscious automata > Machine learning |
| Divisions: | School of Computing > Master of Science in Data Analytics |
| Depositing User: | Tamara Malone |
| Date Deposited: | 26 May 2023 16:43 |
| Last Modified: | 26 May 2023 16:43 |
| URI: | https://norma.ncirl.ie/id/eprint/6669 |
Actions (login required)
 |
View Item |