Health tracking with smartwatches or fitness trackers for personalized health management and self-optimization has become increasingly popular. Today, around 260.7 million users track their steps, heart rate, stress levels and other parameters on a daily basis (Statista Health Market Insights, 2024). However, many of these self-tracking solutions rely on invasive devices that require direct skin contact and are often high in cost. A promising alternative is remote health tracking via camera, which could open up new possibilities. For example, a health tracker integrated into the computer camera could be synchronized with a digital calendar, allowing meetings to be scheduled and rescheduled based on the current stress level.

AI-based remote photoplethysmography algorithms are an innovative approach that enables contactless health monitoring using standard, low-cost cameras. A critical step in this process is the identification of specific areas of the face, known as region of interest (ROI), such as the forehead or cheeks. Stable tracking of the ROI is essential for extracting accurate and reliable heart rate signals. However, influencing factors, such as different lighting conditions, head movements, camera angle and position, make it difficult to obtain reliable measurements in real-world conditions. 

The aim of this thesis is to investigate and compare open source methods for real-time face and ROI recognition. First, common frameworks will be identified through a systematic literature review, and then a prototype will be implemented to evaluate them under selected influencing factors.

Statista Health Market Insights (2024). Statista Health Market Insights. de.statista.com/statistik/daten/studie/1460774/umfrage/nutzer-von-fitnesstrackern-weltweit/. Retrieved 08.05.2025

The digital health market is increasingly moving from a niche market to a mainstream market, and is expected to grow at an annual growth rate of 6.88% to reach a projected market volume of USD 258.25 billion by 2029 (Statista Market Insights, 2024). Health monitoring is an important sub-segment within the digital health market. Smartwatches or smart rings that monitor heart rate or other fitness metrics are already widely used by people for various reasons.

However, these devices have certain drawbacks as they require direct skin contact and are often very expensive. As a result, modern solutions for contactless and cost-effective health monitoring have gained considerable attention in recent years. In particular, new potential is emerging in areas such as road accident prevention and telemedicine, where non-invasive solutions are particularly valuable.

Recent advances in artificial intelligence have significantly improved the accuracy of remote photoplethysmography (rPPG) algorithms. Using these algorithms, heart rate and other vital signs can be measured using a standard RGB camera, enabling completely contactless health monitoring.

The aim of this work is to develop and validate an AI-based rPPG algorithm for real-time heart rate extraction. A prepared test dataset (UBFC-Phys) will be used to train and develop the algorithm. In addition, a small data sample will be collected using a reference measurement device (e.g. ECG chest strap) for validation purposes. The resulting data will then be compared and evaluated using selected performance indicators (e. g. mean absolute error (MAE), Pearson correlation coefficient).

Statista Market Insights (2024). Digital Health. Statista. www.statista.com/outlook/hmo/digital-health/worldwide Retrieved 08.05.2025

Implementing artificial intelligence (AI) models for medical tasks like diagnosis or prognosis promises to support medical staff in their decision making and improve the overall quality of healthcare. However, in order to achieve effective usage of AI for decision support, some potential problems have to be resolved. These include issues of trust, overconfidence, and legal requirements. A popular approach to make AI models more trustworthy, transparent, scrutable, and generally understandable lies in AI explanations. The research community of explainable AI has developed a wide array of methods that attempt to extract valuable insight about the reasoning of any given AI model. While most scholars have developed methods that attribute importance values to individual features to indicate their significance for a given prediction or for the global model behavior, others have taken inspiration from the social sciences and tried to construct more intuitive, human-like explanations. Among these are counterfactual explanations, also known as contrastive explanations. These provide alternative sets of minimally changed inputs that lead to a different model output. Presenting diverse counterfactuals enables insight into the model’s reasoning process in a different way to attribution-based approaches.   

This Master thesis project consists of a systematic literature review (SLR) on contrastive explanations for AI models in the field of medicine. Based on this SLR, the student identifies requirements for the development of an interface that provides contrastive explanations for a specific medical task. Expert interviews with medical professionals will be conducted, recorded, transcribed, and analyzed (e.g., via tools like MAXQDA) to evaluate the developed interface.

Among the high-stakes decision-making contexts that use artificial intelligence (AI), finance is one of the fields that sticks out. However, applications such as fraud detection, credit scoring, and stock price forecasting still require insight into the black box of modern deep learning models. Even if poor decisions in finance, for instance, due to bias or poor data quality, do not directly harm people, they can negatively impact human well-being. Hence, AI explanations to foster trust and improve decision making are required. We intend to research explainable AI (XAI) in finance, which involves an analysis of use cases, methods, and previous experimental evaluations. 

To develop a better understanding of XAI in finance, the Bachelor student carries out a systematic literature review (SLR). This SLR is followed by a series of expert interviews to assess the current state of AI and XAI usage in the financial industry. The interviews must be recorded, transcribed, and analyzed (e.g., via tools like MAXQDA).

The literature on artificial intelligence (AI) explanations comprises two primary explanation methods: attribution-based and counterfactual-based. Through the differences in these approaches, two criteria for good explanations are optimized: necessity and sufficiency. Methods looking for counterfactual explanations elicit necessary features, while methods that look at feature attribution focus on sufficient feature values. Mothilal et al. (2021) propose a framework unifying both methods to evaluate the different approaches with respect to those two criteria for good explanations. Research into metrics for evaluating explanations is relevant because, unlike most prediction and classification tasks, there is no ground truth to evaluate the correctness or quality of explanations. Mothilal et al. (2021) rely on three datasets from the credit-scoring domain and a case study on hospital admission to test their framework. We intend to build on this study and examine, whether the results presented by Mothilal et al. (2021) transfer to different datasets and explanation techniques. 

This Master thesis project builds on previous work by reviewing novel methods for attribution-based and counterfactual-based explanations from the literature, applying these to a new selection of datasets from the medical domain, and evaluating whether more recent approaches to AI explainability better fulfill the criteria of necessity and sufficiency. 

Reference:

• Mothilal, R.K., Mahajan, D., Tan, C., & Sharma, A. (2021, July). Towards unifying feature attribution and counterfactual explanations: Different means to the same end. In Proceedings of the 2021 AAAI/ACM Conference on AI, Ethics, and Society (pp. 652-663).

Modern smartwatches and smart rings increasingly rely on photoplethysmography (PPG) algorithms to measure vital signs such as heart rate, sleep patterns and stress levels. These sensors are particularly attractive because they are non-invasive and inexpensive, which makes them suitable for integration into everyday life. In recent years, advances in signal processing and machine learning have significantly improved the ability to extract meaningful insights from raw PPG data. For example, the first commercially available smartwatches are now able to measure blood pressure, offering an exciting solution for health monitoring. However, most commercial consumer devices do not allow researchers and individuals to access the raw PPG signal, significantly limiting the potential for innovation.

The aim of this thesis is to first define the potential of PPG-based sensors through a systematic literature review. Subsequently, a prototype will be developed that enables the streaming of raw PPG data in real time.

Note: A PPG sensor is provided for technical implementation.

Smart camera-based monitoring systems are capable of tracking and monitoring human activities. In recent years, systems have been developed that enable contactless monitoring of vital signs such as heart rate and oxygen saturation, capture human behavior such as posture and eye activity, and are able to detect human emotions. In contrast to other monitoring systems, cameras have the advantage of being low-cost and unobtrusive, which facilitates their implementation in different environments. Despite the technological capabilities of these systems, their use in the real world is currently limited. 

Therefore, the aim of this Bachelor thesis is to identify use cases and their potential in real-life scenarios. In order to successfully complete this Bachelor thesis, use cases will first be identified through a systematic literature review and then their potential will be explored through an online survey (e.g. LimeSurvey).