Presentation PFE
Download as PPTX, PDF0 likes1,614 views
This document presents an application developed for detecting lung cancer using AI by analyzing x-ray and histopathological images. It covers essential theoretical concepts such as object detection, convolutional neural networks, and machine learning models, while also detailing the architecture and training processes utilized in the project. The conclusion emphasizes the educational value gained from the project, particularly in understanding the seriousness of lung cancer and collaboration in technology development.
Presentation PFE
- 1. LUNG CANCER DETECTION 1 CADI AYYAD UNIVERSITY HIGH SCHOOL OF TECHNOLOGY SAFI COMPUTER SCIENCE Realized by: Mohamed -Hamza Ait Benaissa Mohamed Boussardi Chaimae Gassir Ayoub Essadeq Supervised by: Mr. Alaoui Fdili Othmane Tutored by: Gamal Mohamed
- 3. Introduction According to a recent study, Doctors spend 75% of each patient visit on consulting electronic medical records, and they may made errors predicting them that’s why we include AI in medical filed because it help to increase the levels of accuracy in detecting and diagnosing disease. In this presentation , we will present an application that can detect lung cancer by selecting an X-ray or histopathological image. 3
- 5. Theorical Concepts : object detection 5 - What is Object detection ? - How we can achieve this operation ?
- 6. Theorical Concepts : CNN 6 • CNN is a deep learning algorithm. • Convolutional neural networks are generally composed of the following layers: Convolution Max-pooling Fully-connected
- 7. Theorical Concepts : input 7 How computer see an image? • Machine sees the image as a table of numbers between 0 and 255. • The numbers are called pixels .
- 8. Theorical Concepts : convolution 8 • Extract the high-level features such as edges from the input image. • Image multiplicate by a filter.
- 9. Theorical Concepts : max pooling 9 • Uses the maximum value from a cluster of neurons at the prior layer. • Reduces the dimensions of the image.
- 10. Theorical Concepts : machine learning 10 • Supervised learning: train the algorithm using data which is well “labeled”. • Unsupervised learning: analyze and cluster unlabeled data.
- 11. Models
- 12. Models : project I 12 Data • This Project is using a Kaggle dataset That contains five classes : - colon benign - colon malignant - lung benign -lung malignant ACA -lung malignant SCC Data & Tools Tools • Pytorch • Google colab • spyder
- 13. Models : project I 13 1. Image Preprocessing 2. Histopathological lung vs Others 3. Lung benign vs malignant classification 4. Please insert a Histopathological lung image 5. Lung malignant type classification Architecture
- 14. Models : project I 14 CNN 1 - 2 convolutional layers (Kernel 5x5) - 2 Max Pool (Kernel 4x4) - 3 convolutional layers (Kernel 4x4) - 1 convolutional layers (Kernel 5x5) - 1 Max Pool (Kernel 4x4) CNN 3 CNN 2 - 2 convolutional layers (Kernel 5x5) - 2 Max Pool (Kernel 2x2) CNNs
- 15. Models : project I 15 Training Loss Validation Loss Training Error Validation Error Execution time(s) CNN 1: Lung vs. Colon 0.043 0.074 0.017 0.024 1317.23 CNN 2: Lung Benign vs. Malignant 0.024 0.027 0.010 0.012 3532.80 CNN 3: Lung SCC vs. ACA 0.098 0.355 0.033 0.106 3941.86 Training
- 17. Models : project II 17 VGG 16 Architecture • Architecture • Modification • Tools : keras & tensrflow • Dataset : kaggle • 2 categories : normal / peneumonia
- 18. - Load the training model. - Take an image from the testing set. 18 Models : project II Testing
- 19. Models : project II 19 Output
- 20. Application
- 21. conclusion 21 We have learned a lot of things in this project: • How dangerous lung cancer is. • New technologies and domain. • How to work in groups. • Any problem has a solution. But generally, we achieved our goals and we developed our application where detect lung cancer and returns the results.
- 22. 22 Any questions? Thanks for your Attention :)