Zobraziť v EDS

Development of a Web-based Skin cancer diagnosis using CNN for multi-level classification of dermoscopic Images

Uložené v:
Podrobná bibliografia
Názov: Development of a Web-based Skin cancer diagnosis using CNN for multi-level classification of dermoscopic Images
Autori: Ifty, Hamim Hasan, Liyanage, Indunil Maheshika, Wijayarathna, Kananka Hewage Shyama Harshani
Rok vydania: 2026
Zbierka: Theseus.fi (Open Repository of the Universities of Applied Sciences / Ammattikorkeakoulujen julkaisuarkisto)
Predmety: Option of Web Development, fi=Tieto- ja viestintätekniikka|sv=Informations- och kommunikationsteknik|en=Information and Communications Technology, skin cancer, neural networks (information technology), machine learning, diagnostics, skin diseases, deep learning, diagnosis, melanoma, artificial intelligence, image analysis, JavaScript, Python, Degree Programme in Information Technology
Popis: This research focuses to develop web-based system for skin cancer diagnosis with the use of Convolutional Neural Networks (CNN), multi-level classification of dermoscopic images. Skin cancer is a serious health concern from around the globe and early detection is important in successful treatment. The proposed system uses the pre-trained CNN models for dermoscopic image classification into seven types of skin cancers with great accuracy. The models are fine-tuned using a large number of images (10015) of dermoscopic images and data augmentation techniques are used to enhance the generalisation of the models. The developed web portal helps the users to upload the images of skin lesions (dermoscopic image) and receive the predicted diagnosis with a maximum of three levels of confidence. The system provides a traceable, inexpensive and accessible solution to early detection to skin cancer, especially where specialized medical expertise is limited. The development of CNNs has shown the potential of CNNs in the area of medical image analysis and has opened new direction for research and development in the field of dermatology. The frontend application was created as the single page application based on react and TypeScript which is a responsive and simple to use interface with tools to upload an image and view the diagnosis. It was implemented with the use of Material UI to ensure a consistent design, Clerk to make sure the user is authenticated, and React hooks to manage state efficiently. It has major characteristics such as real time prediction display with confidence score, automated generation of PDF reports, the history of analysis and a fully responsive design to desktop and mobile devices.
Druh dokumentu: bachelor thesis
Jazyk: English
Relation: http://www.theseus.fi/handle/10024/910399
Dostupnosť: http://www.theseus.fi/handle/10024/910399
Rights: fi=All rights reserved. This publication is copyrighted. You may download, display and print it for Your own personal use. Commercial use is prohibited.|sv=All rights reserved. This publication is copyrighted. You may download, display and print it for Your own personal use. Commercial use is prohibited.|en=All rights reserved. This publication is copyrighted. You may download, display and print it for Your own personal use. Commercial use is prohibited.| ; Avoin saatavuus / Open access / Öppen tillgång ; open access
Prístupové číslo: edsbas.6CB5BE65
Databáza: BASE
Popis
Abstrakt:This research focuses to develop web-based system for skin cancer diagnosis with the use of Convolutional Neural Networks (CNN), multi-level classification of dermoscopic images. Skin cancer is a serious health concern from around the globe and early detection is important in successful treatment. The proposed system uses the pre-trained CNN models for dermoscopic image classification into seven types of skin cancers with great accuracy. The models are fine-tuned using a large number of images (10015) of dermoscopic images and data augmentation techniques are used to enhance the generalisation of the models. The developed web portal helps the users to upload the images of skin lesions (dermoscopic image) and receive the predicted diagnosis with a maximum of three levels of confidence. The system provides a traceable, inexpensive and accessible solution to early detection to skin cancer, especially where specialized medical expertise is limited. The development of CNNs has shown the potential of CNNs in the area of medical image analysis and has opened new direction for research and development in the field of dermatology. The frontend application was created as the single page application based on react and TypeScript which is a responsive and simple to use interface with tools to upload an image and view the diagnosis. It was implemented with the use of Material UI to ensure a consistent design, Clerk to make sure the user is authenticated, and React hooks to manage state efficiently. It has major characteristics such as real time prediction display with confidence score, automated generation of PDF reports, the history of analysis and a fully responsive design to desktop and mobile devices.