The thesis was related to detection of anxiety and depression using low density EEG, that is, 8 channels EEG and 1 Channel Ear EEG. The work involved simultaneous detection of anxiety and depression and hence is a multiclass problem. The work was challenging because it not only required knowledge in signal processing and the machine learning but also required understanding of the mental health disorders. The complexity of the project increased due to the multi-level classification, because of similarity of anxiety with depression in terms of neural mechanisms. The student optimized various deep learning models to achieve this objective and tested it on three public datasets. The results are promising.

The thesis is well written and contains content which is publishable as a conference paper. The student is encouraged to prepare the draft for the paper.

The student had weekly meetings and he came to meetings on time and he regularly attended the meetings. I found him to be punctual and hardworking student. He faced many challenges and worked hard to overcome the challenges. He was very active and quite focused on his project.

Evaluation level: zadání téměř splněno

The assignment requirements have been fully fulfilled. The thesis successfully addresses all objectives. A minor reservation concerns the limited discussion of model explainability and interpretation of the learned features, which could have provided additional information to the decision-making process of the proposed deep learning models. However, that does not affect the overall thesis because that was not the primary assignment of the thesis.

Overall thesis is logically structured, nicely written, and easy to follow. The chapters are in a coherent manner and provide detail for understanding the proposed solution. Minor reservations about the 2nd chapter, which could have a summary table comparing related studies, and absence of a visualization showing the selected EEG electrode configuration.

The thesis uses relevant literature closely related to the topic. Deep learning approaches and recent EEG studies are discussed in literature chapter. A minor point I want to raise in the literature review, is comparative summary table of the reviewed studies and a broader discussion of traditional machine learning approaches alongside deep learning methods. 

The thesis provides original experimental results, including the comparison of multiple deep learning architectures and in-ear EEG for anxiety and depression. The binary classification results are good enough, while the four-class classification task still needs further improvements. The presented work provides a strong base for future work and could potentially lead to a publishable research contribution if complemented with additional aspects such as explainable AI, uncertainty estimation.