Collaboration grants
Deep learning for the prediction of prosthetic loosening in the hip (Anders Eklund, IMT, and Jörg Schilcher, RÖ and BKV)
Total hip replacement (THR) has revolutionized hip disease treatment, earning the title "surgery of the century."
In Sweden, the prevalence of THR in individuals over 65 is 8 %, in those over 85, 14 %. The main reasons for reoperations after THR are loosening and prosthetic joint infection, leading to approximately 2,200 reoperations annually in Sweden. Prosthetic loosening is the single most frequent reason for revision surgery (exchanging an old prosthesis with a new one), accounting for approximately 21 % of cases. Prosthetic loosening causes loss of function, pain, and loss of independence for patients, resulting in large costs for society.
In this project we therefore propose to use artificial intelligence (AI) to predict prosthetic loosening before it happens, and thereby minimize suffering for patients by rapidly detecting when reoperation is needed. This will be achieved using Sweden's unique combination of high-quality health care registers and digitized radiographs, making it possible to export radiographs from previous time points for all patients who have undergone revision surgery that can be used to train prediction models.
Health literacy in digital interventions for foreign-born women: an overlooked opportunity to aid health? (Pontus Henriksson, HMV)
Health among foreign-born women, including reproductive health, is an important priority for public health. Previous research, including our own, has shown great differences in the risk for pregnancy complications, obesity, and underweight between different countries of birth and regions. This motivates efforts to reduce inequality in reproductive health.
Health literacy is people’s knowledge, motivation, and ability to access, understand, evaluate, and implement health information for judgement and decisions in everyday life regarding health care, preventive care, and health-promoting work with the purpose of maintaining or improve quality of life. e-health literacy is health literacy specifically for electronic sources. Poor health literacy has consistently been related to worse health results, increased mortality risk, and reduced quality of life. Previous research has shown that low health literacy and low e-health literacy is more common among foreign-born people, which can lead to worse health results and reduce the efficacy of digital health interventions.
Currently, there is a lack of digital interventions that aim to improve health literacy among foreign-born women, and a first step is to do research to develop culturally adapted digital interventions. It is also unclear to what extent health literacy is crucial for the ability to benefit from digital lifestyle interventions.
In study 1, we will perform a qualitative study to investigate foreign-born women’s experiences of accessing, understanding, evaluating, and use health information. Thereafter, we will investigate the needs and wishes regarding digital interventions that can support their health literacy regarding, e.g., structure, functions, and contents. Semi-structured interviews will be performed with about 20 Arabic- and Somali-speaking women.In study 2, we will investigate whether health literacy affects the efficacy of a digital intervention for diet quality and physical activity among foreign-born women after pregnancy. We will use data from a randomized controlled study named PRIMI (Promoting Reproductive health In MIgrant women) which investigates the effect of a digital intervention on lifestyle among 200 foreign-born women who recently have given birth.
The project includes collaboration between researchers at Linköping University (Pontus Henriksson, Ulrika Müssener, Marie Leksell, Maryam Shirvanifar, Anna Seiterö, Aisha Salah Ahmed), researchers at Karolinska Institute (Josefin Wångdahl, Viktor Ahlqvist, Daniel Berglind) and personell at Flyktningsmedicinskt centrum within Region Östergötland (Baydaa Al-Saedi, Malin Creutz, Emira Bajric). This collaboration will be crucial for the ability to perform the research project, but we also see a very large potential to improve the collaboration to further promote healthy lifestyle and health among people who have immigrated to Sweden.
Novel Application of Large Language Models for Patient Safety: Using AI to Make MRI Scans Safer by Detecting Unknown Medical Implants (Peter Lundberg, RÖ and HMV)
Each year, around 30,000 MRI scans are conducted in Östergötland and a growing number of these patients, about 20–25 %, have implanted medical devices like pacemakers or joint replacements. While MRI scans are a powerful diagnostic tool, they can pose serious risks to patients with metal implants. These devices can behave like unintended antennas during an MRI, absorbing electromagnetic energy and potentially causing severe or even life-threatening injuries.
The problem? Healthcare systems store information about implants in digital medical records, but the data is usually written in unstructured text, making it hard to find or interpret quickly. Currently, identifying whether a patient has an implant sometimes involves a slow, manual process where multiple experts must carefully review a patient’s full medical history.
A new research initiative led by Region Östergötland and Linköping University’s AIDA Data Hub aims to solve this problem using AI. Our goal is to build a language-independent system that can automatically detect mentions of implants in electronic medical records, even when written in complex or inconsistent language. The ultimate aim is to flag any potential risk before a patient is sent into an MRI scanner. To power this solution, the team plans to fine-tune an advanced AI large language model called XLM-RoBERTa. This model, based on Google’s BERT and initially trained on 100 languages, will be adapted to handle the unique challenges of medical language, including technical terms and fragmented grammar.
However, training such a specialized model requires considerable computing power and also access to vast amounts of sensitive patient data. The project will rely both on in-house computing and on compute infrastructure provided by AIDA Data Hub to efficiently process and analyze the data. To ensure privacy and comply with strict data protection laws, the researchers will use federated learning, a method that allows models to be trained on data from multiple sources without actually sharing the data itself. The project also aims to explore the legal and technical implications of using cloud-based computer platforms to handle sensitive health data. If successful, it could lead to the creation of a working prototype of the AI implant-detection system that could be scaled up for wider use in both public healthcare and private medical tech industries.
In addition to the Swedish partners, the team is collaborating with researchers at the University of Copenhagen and Norway’s National Centre for E-health Research in Tromsø. This Nordic collaboration will help improve the AI’s accuracy and adaptability by learning from more diverse data sources across different healthcare systems. More broadly, this research supports the European Union’s growing ambitions to make health data more accessible and usable for both clinical care and innovation. While patient safety during MRI scans is the starting point, the long-term vision includes unlocking the hidden potential of unstructured medical data to transform healthcare delivery and research across Europe.
Seed grants
Co-design of a chatbot based on Large Language Models for informal carers in a health care context (Hanna Allemann, HMV)
I’m very grateful for the support from Area of Strength e-Health. It gives me as a junior researcher some wind beneath my wings, and the grant helps me to achieve more independence as a researcher. With aid of previous support from Area of Strength e-Health, an AI-based chatbot prototype has been developed. The prototype integrates a Large Language Model (LLM) with context-specific information from a co-designed support program for informal carers of people with heart failure. LLMs have previously been shown to be able to give correct and relevant advice with a person-centered tone. Digital solutions based on these LLMs can also offer more personalized solutions adapted to the individual’s needs and abilities.
The aim of the new project is to, with aid of co-design, refine, design, and study the usage of the chatbot. I want to ensure that it adheres to the laws, regulations, and values of healthcare as well as the view of the informal carers. Thus, our intention is to include informal carers, healthcare professionals, technicians, and researchers in the process. This is important to study thoroughly since there are risks with AI. Some risks that often are mentioned are that AI can give answers that include prejudices or falsehoods. In some circumstances, this can have a negative effect on the user. Some users may also trust the AI too much, which can be problematic. This, or completely mistrusting the AI, reduces its usefulness. These aspects will also be investigated in this project.
Evaluation of the LVAD driveline bandage change through remote-controlled video observation: Impact on infections and hospitalizations (Magda Eriksson-Liebon, HMV)
Heart failure is a serious condition affecting over 200 000 persons in Sweden. The condition causes reduced quality of life, frequent hospital visits, and high health care costs. Since access to donor hearts is limited, mechanical circulation support such as the Left Ventricular Assist Device (LVAD) is used to a greater and greater extent as treatment for advanced heart failure. The LVAD aids the left ventricle with pumping blood and can function both as a bridge to transplant and as destination therapy for patients that are not feasible for heart transplantation. The purpose of LVAD treatment is to prolong survival and improve the patient’s health-related quality of life. However, living with an LVAD requires a great deal of personal responsibility for self-care, especially regarding handling the driveline – the cable that connects the pump inside the body with control system and power source outside the body. Infection of the driveline is one of the most common and most serious complications and a common cause of re-hospitalization. It is thus crucial that both patients and relatives have good knowledge of aseptic techniques, bandage change, and daily driveline care.
The grant from e-Health will be used for a project studying the possibility of using remote-controlled video recordings to evaluate self-care at home. By filming the bandage change in the patient’s natural environment, without the presence of health care staff, conditions for more authentic assessment of how the self-care actually is performed are created. The recordings will subsequently be reviewed by health care staff with specialist competence in LVAD care. Based on the results of these video observations, the next step in the project will be to develop a digital education program for patients and relatives with a focus on practical training in driveline care. In a subsequent pilot study, we plan to evaluate the effect of the program on clinical results such as driveline infections and unplanned re-hospitalizations. By using video recordings as a method to observe driveline care in the home, the study can aid the development of an aimed education program that strengthens the ability of patients and relatives to prevent infections and improve self-care during LVAD treatment.
Development of an online-based resource for health care providers interacting with people with acquired communication disorders in clinical settings: tools for patient inclusion (Sophia Lindeberg, BKV)
The aim of this project is to investigate the potential of a digital resource to improve the communication between health care staff and people with acquired communication disorders (disorders due to illness or injury). The focus will be on identification of obstacles and resources in clinical meetings and digital contacts. Effective communication between health care staff and patients is essential for personalized care. People with communication disorders, such as aphasia (difficulties in speaking or understanding languages) or dysarthria (difficulties with articulation), often experience challenges in interacting with people outside their closest acquaintances. Health care staff also experience considerable communication obstacles when meeting these patients. During recent years, the importance of conversation partners has been emphasized. Programs for conversation partner training has been shown to improve the communication between people with communication disabilities and their acquaintances. However, the effectiveness of such training in complex settings, such as in health care, is less well investigated.
Implementation of iCBT for mental health in heart patients (Johan Lundgren, HMV)
Problems with mental health are common in people with heart disease. Many patients with heart disease experience stress, anxiety, and depression connected to their heart disease and life situation. Despite how common these problems are, there is currently a lack of structured and well-functioning treatment strategies that are extensively implemented clinically to meet the needs of these patients. The aim of the project is to investigate the conditions for implementation of internet-based cognitive behavioral therapy (iCBT) to relieve psychological distress among patients with heart disease. The research group wants to establish collaboration with clinical units within Region Östergötland to facilitate the initiation and development of a larger research project about this implementation. Specifically, the project wants to identify and analyze factors that facilitate or hinder implementation of iCBT within heart care, including organization aspects and resource needs.
The project is based on the theories of implementation science and focuses on identification of potential obstacles and opportunities on organizational level for the introduction of iCBT within routine heart care. The study is using qualitative design with co-creation, where the participants are recruited from both primary and specialist care. We plan to use 6 – 8 focus groups with 5 – 7 participants per group, including different professions such as nurses, medical doctors, physiotherapists, and management.
The participants in the focus groups will identify factors that facilitate or hinder implementation of iCBT as well as strategies to overcome the obstacles. To ensure that the perspectives of the clinical staff are clearly included, the analysis will be performed together with part of the clinical staff in the focus groups with help of the Framework Method. The project is expected to start in the third quarter of 2025 and data collection is planned to continue until the beginning of 2026. The results from the study will contribute to a future more extensive implementation project and aid the integration of iCBT as a routine part of heart care, which may improve the care for patients with heart problems and psychological stress. The project aims to create a more person-centered close care and improve the results for individuals who live with heart disease.
Distributional cost-effectiveness of a digital smoking cessation intervention (Katarina Ulfsdotter Gunnarsson, HMV)
The number of smokers have decreased in the last decades but is still one of the greatest health risks. A randomized controlled study that was carried out 2022 in Region Östergötland showed that the chance to stop smoking was doubled for those who had access to a digital smoking cessation intervention compared to those who got standard support. However, the results showed considerable variation within the study population. For example, heavy smokers did not get as positive results as light smokers, and the effect was greater among men. Since the effects of the intervention are heterogeneous, it is uncertain if the intervention is cost-effective for all users.