【113-2微學分課程】AI醫學影像分析實作(英文授課)

AI for Medical Image Analysis(EMI)

Course Overview and Objectives：

This course explores the practical applications of artificial intelligence in medical image analysis, focusing on deep learning techniques, image preprocessing, model training, and evaluation. Participants will gain hands-on experience with AI models for various imaging modalities, including MRI, CT, X-ray, Histopathology, and retinal scans. The course will also cover key challenges such as model interpretability, bias mitigation, and regulatory considerations in medical AI.

By the end of the course, participants will:

  1. Understand the fundamentals of different types of medical images and AI-driven medical image analysis, including deep learning architecture and preprocessing techniques.
  2. Develop proficiency in implementing and fine-tuning AI models for different medical imaging tasks.
  3. Evaluate model performance using appropriate metrics and interpret AI-generated results.
  4. Identify and address challenges related to data bias, fairness, and regulatory compliance in medical AI applications.

Gain hands-on experience with AI frameworks and tools commonly used in medical image analysis.

前往選課系統

授課教師

電子研究所 賴伯承 教授、腦科學研究所 紀以柔 博士生、電機資訊國際博士學位學程 冉恩達 博士生

對應總課程名稱

人工智慧應用與實作 Artificial Intelligence Applications and Implementations

課程日期

3/22 13:00-16:00 PM 3/23 13:00-16:00 PM 3/29 13:00-15:00 PM

課程總時數

8小時

上課地點

交大校區EA101

修課人數

15人

先修科目或先備能力

Please complete the pre-course survey: Google form

自備物品

laptop

課程教材

1. A. Sweigart, Automate the Boring Stuff with Python, 2nd Ed., No Starch Press, Dec. 2019 2. Literature reading. 3. Online resources.

作業、考試、評量

Labs and homework (done individually) 20%
Final project (done in groups of 1-3 members) 80%

Topic Content outline Teach Demonstration Exercises Others
Principles and Diagnosis Overview Overview of medical imaging modalities and principles, introduction to example diseases (autism, pneumonia, dementia, diabetes mellitus…) in class 1hr 0hr 0hr 0hr
Data preprocessing Focus on how to preprocess with Python to make into Pytorch dataloader Explain two examples: MRI and Histo Data preprocessing lab: give each team a dataset and prepare a pytorch dataloader (will use later for group project) 0.5hr 0.5hr 1hr -
Computer Vision with Deep Learning Introduction This session covers the fundamentals of computer vision and deep learning, including structures of models, feature extraction, image classification, and hands-on implementation of medical image analysis models. 0.5hr 0.5hr - -
Image Classification Colab Tutorial This session provides a hands-on tutorial using Google Colab to implement image classification models, covering data preprocessing, model training, evaluation, and visualization of results. - 0.5hr 0.5hr -
Results Interpretation and Reasoning This session focuses on interpreting AI model results by integrating medical knowledge of diseases and clinical contexts, analyzing performance metrics, visualizing decision-making processes, and making informed inferences to support clinical decision-making. 0.5hr 0.5hr - -
Group project presentations This session features group project presentations where participants apply and integrate the skills learned throughout the course, showcasing their end-to-end AI model development, from data preprocessing and training to result interpretation and clinical reasoning. - - 2hr -
日期 課程進度、內容、主題
03/22 1.1 Introduction to medical imaging modalities (MRI, CT, X-ray, Retinal Imaging, Histopathology) 1.2 Fundamental principles behind different imaging techniques 1.3 Common medical conditions analyzed through imaging (e.g., tumors, cardiovascular diseases, retinal disorders; autism, pneumonia, dementia) 1.4 Role of AI in medical image diagnosis and clinical decision-making 2.1 Importance of data preprocessing in AI-driven medical imaging 2.2 Image normalization, augmentation, and enhancement techniques 2.3 Handling missing data and noise reduction methods 2.4 Dataset loading, preprocessing, and augmentation in Python
03/23 3.1 Fundamentals of computer vision and its applications in healthcare 3.2 Overview of deep learning architectures for medical imaging (CNNs, transformers, hybrid models) 3.3 Feature extraction and representation learning in medical image analysis 3.4 Challenges and limitations of deep learning in clinical practice 4.1 Hands-on implementation of image classification models using Google Colab 4.2 Training deep learning models for disease classification (e.g., normal vs. abnormal scans) 4.3 Performance evaluation with key metrics (accuracy, precision, recall, AUC-ROC) 5.1 Understanding AI model predictions and performance metrics 5.2 Visualization techniques (Grad-CAM) for explainable AI in medical imaging 5.3 Integrating clinical knowledge with AI results for diagnosis and decision-making 5.4 Ethical considerations and bias detection in medical AI applications
03/29 6.1 Presentation of team-based AI projects on medical image analysis 6.2 Demonstration of end-to-end model development, from preprocessing to interpretation 6.3 Peer review and feedback session to improve clinical applicability of AI models 6.4 Discussion on challenges, lessons learned, and future improvements

常見問題