Anomaly Detection for Mental Health Conditions

Project description

Mental illnesses, including depression, anxiety, bipolar disorder, and schizophrenia, are increasingly recognized as significant global health challenges. According to the World Health Organization (WHO), mental health disorders account for a large share of the global disease burden, contributing to disability, reduced quality of life, and increased healthcare costs. These diseases typically develop over long periods, often with few or no symptoms in the early stages, which makes early detection difficult. As a result, they are often diagnosed at more advanced stages, when treatment options become more limited, less effective, and far more costly. The conventional healthcare model remains largely reactive, focusing on the treatment of mental illnesses once symptoms manifest, rather than addressing early warning signs that could lead to prevention. This treatment-first approach is often less effective because it tends to manage the symptoms and complications of the disease rather than the underlying causes or risk factors. Furthermore, existing methods do not maintain the health trajectories of individuals. Early diagnosis and timely intervention, on the other hand, could significantly reduce the burden of mental illnesses by halting or slowing disease progression, improving quality of life, and cutting healthcare costs. However, identifying those at risk early enough for preventive measures to be effective remains a major challenge in clinical practice. Recent advancements in digital health technologies, such as wearable devices, electronic health records (EHRs), and mental health tracking apps, have created new opportunities for more proactive strategies. Wearable devices can capture continuous biometric data such as heart rate variability and sleep patterns, while mental health tracking apps monitor mood, stress, and emotional states. EHRs provide longitudinal data, including medical histories and psychiatric assessments, offering rich insights into an individual’s mental health over time. However, while the availability of these diverse data sources has grown, healthcare systems are still grappling with how to leverage them for early intervention. Manually interpreting large, heterogeneous datasets to identify subtle patterns that indicate early signs of disease is beyond the capability of human clinicians alone, which is where artificial intelligence (AI) and machine learning (ML) techniques come in. An individual can be in many health trajectories, and anomalous space modelling can be used to model these trajectories progressively to find deviations from an individual’s normal health trajectory—before those deviations lead to clinical symptoms. Anomaly detection is a branch of AI focused on identifying outliers or unusual patterns in data that differ significantly from expected behaviour. In mental illnesses, anomaly detection can monitor health data for subtle shifts in biometric or behavioural markers that precede disease onset. For example, less sleep, varying breathing levels, changes in heart rate, or a decline in physical activity might not warrant immediate concern. However, these health and behavioural data can be mapped on a trajectory to predict possible health concerns. Identifying these anomalies can provide clinicians and patients with valuable early warning signals, allowing them to take preventive actions before the disease progresses. Hence, this research proposes an AI-driven anomaly detection approach that goes beyond detecting single instances of anomalies. The aim is to study the personal health trajectory of individuals, how their health metrics evolve, and identify significant deviations or intersections along this path using anomaly detection techniques. These intersections represent points where health parameters begin to deviate from their normal trajectory, signalling potential risks for disease progression. In other terms, these intersections are the personal health trajectory with the Isomatic path of various diseases. By continuously monitoring these trajectories, it is possible to detect early warning signs and direct individuals to seek medical assistance. The concept of personal health trajectories recognizes that health is not static but rather dynamic, influenced by a wide range of factors that interact over time. A key advantage of this approach is that it allows for the detection of subtle shifts in health long before they manifest as overt disease symptoms. This proactive method could revolutionize mental illness prevention, enabling healthcare providers to intervene earlier and more effectively, leading to better patient outcomes and a reduction in the societal and economic burden of mentally unstable persons. The integration of diverse health data sources—EHRs, wearable devices, mental well-being apps, and lifestyle data—combined with AI-driven anomaly detection represents a significant opportunity to shift healthcare from a treatmentbased model to a prevention-focused approach. Hence, this research aims to harness the power of AI to continuously monitor health data, detect early signs of mental illnesses, and implement personalized interventions that maintain health and prevent disease progression. This transformative approach could reshape how mental illnesses are managed, leading to improved long-term health outcomes and a reduction in healthcare costs worldwide.

Assumed knowledge

1. Programming
2. Database Management

Note: You need to register interest in projects from different supervisors (not a number of projects with the one supervisor).
You must also contact each supervisor directly to discuss both the project details and your suitability to undertake the project.