Pilot study of breast cancer classification using deep learning approach
Breast cancer is harmful diseases and also the most common cancer causes of cancerrelated deaths among women. Traditional diagnosis and detection method requires a lot of experience, professional knowledge, and human sometimes might prone to fall diagnostic errors. The effective way to reduce the mo...
保存先:
| 第一著者: | |
|---|---|
| フォーマット: | Final Year Project / Dissertation / Thesis |
| 出版事項: |
2020
|
| 主題: | |
| オンライン･アクセス: | http://eprints.utar.edu.my/3946/1/17ACB00134_FYP.pdf http://eprints.utar.edu.my/3946/ |
| タグ: |
タグ追加
タグなし, このレコードへの初めてのタグを付けませんか!
|
| 要約: | Breast cancer is harmful diseases and also the most common cancer causes of cancerrelated deaths among women. Traditional diagnosis and detection method requires a lot of experience, professional knowledge, and human sometimes might prone to fall diagnostic errors. The effective way to reduce the mortality rate is to diagnose cancer in the earlier stage by screening. The main objective of this project is to develop a preliminary disease diagnosis system for breast cancer detection and segmentation by using a state-of-the-art deep learning technique. Moreover, the methodology applied in this project is using Mask R-CNN pre-trained model. This project was focused on a fundamental study and developed by customizing and training the pre-trained models with the breast ultrasound images. Thus, the model deployed in this project achieved the lowest loss, which is 0.245. By applying this model, three prediction outputs results will show the class label, bounding box coordination and segmentation mask for the detected object. By implementing this project, Malaysia woman might have the benefit and able to prevent cancer at the earlier stage, and it might reduce the rate of mortality caused by breast cancer. |
|---|