Pisa, 27-31/05/2024.
Prof. Rafael Garcia from Universitat de Girona and coordinator of the iToBoS project has taught the “Mastering Machine Learning Techniques: Application to Melanoma Detection” course in Pisa between 27 and 31 May 2024.
The course begun with an overview of the iToBoS EU project, emphasizing the significant role of machine learning in enhancing the accuracy and efficiency of melanoma detection. This set the stage for the technical content that followed, highlighting the real-world impact of these technologies in healthcare in general, and in the detection of melanoma in particular.
This 8-hour course covered a comprehensive exploration of machine learning techniques applied to medical imaging, leading towards the early detection of melanoma. The course catered to a diverse audience, including both master and PhD students, with a strong interest in machine learning in the medical imaging context.
Participants then were introduced to the foundational concepts of machine learning, starting with linear regression and rapidly transitioning to logistic regression, to focus on classification problems relevant to medical diagnostics.
After this introduction the course delved into artificial neural networks. The progression from logistic regression to neural networks is a natural evolution from simpler linear models to more complex, non-linear models capable of handling a wider range of data and problem types in machine learning.
Continuing from the exploration of artificial neural networks, the course then introduced advanced techniques with a special focus on convolutional neural networks (CNNs) and attention-based mechanisms. These methods are crucial for handling the intricate details necessary for effective medical image analysis, particularly in dermatology.
Participants engaged in practical sessions using dermatological imaging datasets, applying CNNs to detect and classify skin lesions with varying degrees of malignancy. This hands-on experience was vital for understanding the nuances of neural network tuning and optimization in real-world medical applications.
A significant portion of the course was devoted to discussing and implementing modality fusion techniques, which involve integrating multiple types of data (clinical records, demographic information and genomics) to improve diagnostic accuracy. This approach was thoroughly covered, demonstrating various strategies like early and late fusion, and attention-based models to enhance feature extraction and classification performance.
The course also addressed critical evaluations of model performance using metrics such as precision, recall, F1 score, and the area under the precision-recall curve (AUC-PR), as well as sensitivity and specificity. These metrics are essential for assessing the effectiveness of machine learning models, especially in the context of medical diagnostics where the cost of false positives and negatives can be high.
Furthermore, ethical considerations, data privacy, and the impact of machine learning on patient outcomes were discussed. This discussion was aimed at preparing participants to not only develop effective technical solutions but also consider the broader implications of their work in healthcare. By the end of the course, participants had gained a solid understanding of key machine learning concepts and their application in the medical field, particularly in melanoma detection.
All the deails at Università di Pisa.