Abstract:

Soon after computers appeared, AI (artificial intelligence) sparked the curiosity of researchers. Lately, it captured even more attention both in mass media and in the scientific community. It is a technology that creates both concerns and promises of positive changes (better automation, decision making, research, and even creativity).

AI is regarded with such interest because it can be applied in a multitude of scenarios and industries. The healthcare benefits intensively from the implementation of the new generation of AI algorithms. The benefits are two-folded, on one side we have automation and better diagnosis and on the other side, we have the enhanced knowledge and the possibility of novel scientific breakthroughs. For example, AI-based algorithms can be used for setting a diagnostic based on medical images, which sometimes is challenging even for experienced personnel.

The recent advances in machine learning and the increase in the size of the available bioinformatics databases are facilitating scientific advances in genetics and molecular biology. This kind of data is very abundant but also very noisy and hard to explore by humans. On the other hand, the new generation of machine learning algorithms can extract meaningful information from very noisy and unstructured data.

Biomedical engineering is the application of engineering principles, materials, and design concepts to medicine and biology for healthcare purposes. Of high interest is the recent revolution of deep learning and the remarkable results in areas such as biomedical image segmentation and classification, automated extraction of biological relations from biomedical literature, and protein bioinformatics. There is no doubt that mastering the foundations of AI is an important skill for a future biomedical engineer.

Problem-based learning (PBL) and case-based learning (CBL) have demonstrated various benefits to both learners and teachers, such as promoting lifelong learning, teamwork, learner satisfaction, and increased motivation. In PBL, facilitators play a minimal role, while in CBL (also called a guided inquiry approach) the interaction between learners and teachers (facilitators) is tighter and can take place in advance of the learning session so that the facilitators can fix incorrect assumptions and biases early on. The basic underlying idea of both strategies is that one learns best by actively researching the topic of interest. As the students progress, they will most probably live positive emotions such as fulfillment and curiosity. When this occurs, the learning takes place naturally. Both paradigms have been shown to be very effective ways of teaching.

In this paper, we propose an online collection of real-world biomedical engineering problems that introduce the student to the latest methods and algorithms in AI, such as using deep learning for diagnosing health issues from colposcopy images or chest X-ray images and analyzing the potency of HIV-1 antibodies. There are also included open problems that can be solved based on the knowledge acquired by following the already solved problems.

Deep learning is a very vast field, that utilizes specific terminology, methodology, and tools. By using this collection of real-world problems, the student will become proficient in using the Python programming language, Jupyter notebooks, Pytorch, and the correct methodology of training and evaluating a neural network.

Keywords:

Biomedical engineering, bioinformatics, artificial intelligence, deep learning, neural networks, problem-based learning.