Data4SmartHealth 2024
28 November 2024, 10:30 – 12:30, Free University of Bozen-Bolzano, Italy
Medical problems
Psychiatry – Filippo Boschello
AI based predicting tool for suicide attempts
Background
Suicide and suicidal attempts are prevalent among psychiatric patients. The C-SSRS (Columbia-Suicide Severity Rating Scale), the NGARS (Nurses’ Global Assessment of Suicide Risk) and the BSSI (Beck Scale for Suicide Ideation) are the most applied questionnaires to screen patients for suicide risk. Despite their validation, these tools are rarely applied in the real-world clinical setting, as psychiatrists show often reluctancies regarding their accuracy, their implementation in the workflow due to lack of time, the possibility that an over-emphasis on risk stratification may lead to disruptive defensive clinical measures. Usually, in their daily clinical practice clinicians opt for a clinical interview to assess the suicidal risk according to a less structured and more narrative approach.
Research question
We are interested in understanding how the written clinical records reported during the follow-up visits before a suicide attempt leading to a hospitalization could be used to develop a text mining algorithm to predict the risk of suicide.
Aims
Our first aim is to investigate whether the clinical items reported into the written records of suicidal outpatients may enable an AI based algorithm to make a reliable prediction about suicide attempts. As a secondary outcome, we would like to explore the relation between features potentially empowering the algorithm-based predictions of suicide-attempts and items included in the most widely applied suicide risk screening tools and clinical assessment.
Speaker’s short bio: Graduated in Medicine from the University of Padua in 2012. Residency course in Psychiatry and Psychotherapy completed at the University Hospital of Verona in 2018, discussing a thesis entitled: “Multivariate analysis of ROI-based fractal dimension: a neuroimaging study in first episode psychotic patients. Insight from the European Study PRONIA”. Visiting researcher at the Laboratory for translational psychiatry – Klinik für Psychiatrie und Psychotherapie – LMU (München, Germany) from September 2017 to July 2018. Master degree in Forensic Psychiatry completed at the University of Florence in 2023. Author and co-author of poster, oral presentations and publications on peer-reviewed journal in the field of neuroimaging, AHDH, psychopharmacology. Currently working as consultant in psychiatry at the Mental Healthcare Service, Bozen-Bolzano Hospital.
(slides)
Diabetology – Dalia Crazzolara
Can AI support the management of physical activity in people with type 1 diabetes?
Type 1 diabetes is an autoimmune disease in which the destruction of pancreatic cells leads to the loss of insulin production. Consequently, without insulin, glucose fails to be transported from the bloodstream to peripheral cells, to be used as energy, leading to increased blood glucose levels. Therefore, maintaining blood glucose levels within the target range of 3.9-10 mmol/L necessitates individuals with type 1 diabetes to undergo exogenous insulin therapy. Insulin requirement is subjective and is influenced by the amount of carbohydrate intake and physical activity. The daily challenge for people with type 1 diabetes mellitus is maintaining glycaemic levels within targets during and after physical activity. How could AI meet this need?
Speaker’s short bio: Dalia Crazzolara is a Physician specialized in Endocrinology, Diabetes & Metabolism and in Internal Medicine. She works at the Diabetology Service, Department of Internal Medicine at Bozen-Bolzano Hospital where she is responsible for management and treatment in type 1 Diabetes and the use of technology. She is currently attending the PhD course in Medical Science at the Paracelsus Medical University in Salzburg, Austria. She is a member of the ethics committee for clinical trials of the South Tyrolean Healthcare Service.
(slides)
Nuclear medicine – Mohsen Farsad
Integrating AI in PET imaging for brain tumors
The integration of artificial intelligence (AI) in functional neuro-imaging with positron emission tomography (PET) for brain tumors is a rapidly evolving field. AI algorithms, especially those based on machine learning and deep learning, enhance the analysis and interpretation of PET data, leading to more accurate and efficient diagnostics. AI algorithms enhance image reconstruction, improving resolution and accuracy in detecting tumor boundaries. Machine learning models analyze PET data to differentiate between tumor types and predict treatment responses, disease progression and treatment outcomes. AI-driven tools also assist in automating image segmentation, reducing manual effort and increasing consistency. Overall, AI integration in PET imaging offers significant advancements in the diagnosis and management of brain tumors. The use of AI in PET imaging is revolutionizing functional neuro-imaging but there are several open questions and challenges. Therefore, there is the urgent need for ongoing research and collaboration between AI developers, clinicians, and regulatory bodies to fully realize the potential of AI in PET imaging for brain tumors.
Speaker’s short bio: Dr. Mohsen Farsad is the head of the Nuclear Medicine Department at the South Tyrol Health Authority in Bozen-Bolzano, Italy. He is a member of Oncology Study Group of Italian Association of Nuclear Medicine and Molecular Imaging and holds the Italian National Scientific Qualification for Associate Professor. He specializes in advanced diagnostic imaging techniques, including PET-CT and SPECT-CT, which are crucial for early detection and treatment of various pathologies, particularly in oncology. Dr. Farsad’s work involves both diagnostic and therapeutic applications of nuclear medicine, utilizing cutting-edge technology to provide personalized patient care.
(slides)
Orthopedics – Marco Santarelli
AI-driven diagnosis and prediction: potential orthopedic clinical applications
This contribution presents a state-of-the-art of potential applications of artificial intelligence in the orthopedic field; it provides elements and ideas to approach the most captivating questions so far, the limits and problems still to tackle. For example, imaging analysis is pivotal to evaluate differential diagnosis, whereas prediction tools are crucial to optimize therapy outcomes. ML and CNNs could be game-changing in both fields. Potential real clinical applications are fracture/tumor recognition and classification, therapy decision making, postoperative outcome prediction.
Speaker’s short bio: Specialist in orthopedics and traumatology since 2018, he worked in the South Tyrol medical service with a focus on knee and shoulder surgery. ÖÄK sports physician since 2021. After years of experience in clinical and surgical field, he joined the Innovation, Research and Teaching service of the South Tyrol Health Authority. PhD student at Paracelsus Medical University.
(slides)
