Aims

Motivation

The amount of wearables and IoT devices is yearly growing. These devices are finding their way into healthcare applications. The data coming from the sensors of these devices are typically time-series data. For most traditional machine learning algorithms, the raw time series data cannot be processed directly.  Therefore, domain prior knowledge, feature engineering, and feature selection are critical to shaping the raw sensor data into interpretable information for a traditional machine learning algorithm. However, hand-crafted features can only be used in one application domain. Deep  learning (DL) techniques have outperformed traditional methods that rely on hand-crafted features in various domains. Owing to their large parametric space, deep learning techniques can infer relevant features directly from the raw input data.   In this course, the students will learn the theory and intuition behind various deep-learning algorithms. Furthermore, the students will learn how to apply state of the art machine learning models to a variety of challenging healthcare-related applications.

A specific information processing application in health care is medical imaging. It is a key technology in modern medicine and it is highly technology driven. Therefore, the students must understand the physical principles of the major medical imaging technologies such as MRI, CT, US, PET and SPECT and have a basic idea of the different major components of these modalities. They should also understand the reconstruction of images from the measured signals for each modality.  

• The student will learn to use publicly available healthcare-related datasets. (MP1)
• The students will learn feature selection and feature construction techniques. Normalization techniques are also often used here (MK1, MK2, MP1).
• The student will learn to implement machine learning tasks such as classification, regression or clustering tasks. She will apply these techniques to the dataset at hand (MK1, MK2, MI2, MP1).
• The student will learn to evaluate the performance of a machine learning task by looking at a precision-recall curve and inspecting the F-measures. She will also learn to evaluate and reduce the bias and variance of machine learning models (MK1, MK2, MI3, MP1).
• Finally, the student will report the findings of the different ML pipeline elements to the other students and write a paper with the most important findings. (MG2, MG3)
• For each major imaging technique (Ultrasound, Magnetic Resonance Imaging, X-ray Computed Tomography, Positron emission tomography and single photon emission tomography), the student knows the physical principles underlying each modality (MK1), the major components of each modality (MK1), and the basics of the reconstruction of images from the raw signals (MK1).

Previous knowledge

Knowledge of calculus, linear algebra and physics are preferable. The student needs to be able to program in MATLAB or PYTHON.

Is included in these courses of study

• Master of Innovative Health Technology (Leuven) 60 ects.