After surgery or other forms of treatment for cancer or obesity, a hospital bed is often replaced by the comfort of home as soon as possible. While discharge typically marks the end of hospital-based care, it does not eliminate the risk of treatment-related complications.
Fatime Oumar (a PhD candidate at UT) addresses this broader challenge by developing dynamic, AI-driven predictive models to detect complications, such as infections, early signs of respiratory decline, or other physiological decline, before they manifest clinically. Some of these adverse events can occur days, weeks, or even months after treatment, frequently without early warning, and sometimes, too late to recognise and manage effectively. Fatime specialises in biomedical engineering and develops AI-based models for the early detection of complications following cancer or obesity treatment. Part of her research focuses on collecting and analysing continuous vital signs (such as heart rate and respiration) using wearable sensors and personalised prediction algorithms. “We try to pick up the early signs, subtle, silent changes in the body, that might signal something is about to go wrong,” she says.
Making the invisible visible
Fatime is part of the Biomedical Signals and Systems (BSS) group at UT and contributes to RECENTRE, a nationwide research program funded by the 4TU. Her work centers on developing predictive models capable of identifying complications following various types of cancer therapy and bariatric procedures. In this context, she collaborates closely with clinical partners at Medisch Spectrum Twente (MST), Ziekenhuisgroep Twente (ZGT), the Princess Máxima Centre for Pediatric Oncology, and the Helen Dowling Institute (HDI).
In a pilot study, patients undergoing major abdominal oncological surgery at MST were provided with a wearable patch that monitored them continuously from home for a 14-day postoperative period. This small, lightweight device discreetly records physiological signals, including heart rate, breathing rate, and physical activity. “These sensors collect high-frequency data that reflect the patient’s real-time condition,” Fatime explains. “We use this data to train AI models that learn what is normal for each patient, creating a personalised baseline.”
When the system identifies a change, such as a slight but consistent increase or drop in heart rate, it flags it as a potential warning. “Often, these deviations occur well before traditional clinical scores would detect a problem,” she adds. “This early detection could support more timely interventions and improve patient outcomes. With some patients, we are seeing signs six hours earlier than current hospital methods. That difference can be meaningful in a clinical setting.”
Cancer, childhood and care
Another dimension of Fatime’s research focuses on long-term survivorship in childhood cancer populations. In collaboration with the Princess Máxima Centre, she is exploring how longitudinal health data, including lifestyle factors, comorbidities, and treatment-related information, can be integrated to predict risks such as cardiac dysfunction that may emerge years after treatment. The goal is to identify early indicators of cardiac dysfunction that may emerge years after treatment, and to guide both lifestyle adjustments and personalised follow-up strategies.
“We aim to develop adaptive risk prediction models that account for changes in a patient’s health over time,” she explains. “Risk is not static. It evolves in response to lifestyle changes, new diagnoses, and treatment history. By anticipating these changes, we can better support timely and tailored follow-up care for survivors.”
A personal fascination
Fatime’s interest in medical data analysis began during her studies in Türkiye, where she completed both her bachelor's and master's degrees in biomedical engineering. Her bachelor's thesis focused on the detection of diabetic retinopathy using retinal imaging and machine learning. For her master's thesis, she developed an evolutionary algorithm to select the most relevant spectral features from urine-based Fourier-transform infrared (FTIR) spectroscopy data, which were then used in machine learning models to classify endometrial and ovarian cancer. “I was fascinated by the idea that computational models could extract meaningful patterns from complex biological data,” she says. “The PhD position at the University of Twente immediately caught my attention, as it brought together my interests in biomedical data analysis, artificial intelligence, and clinical relevance. It offered a unique opportunity to deepen my expertise while contributing to meaningful advances in healthcare.”
From hospitals to homes
Fatime’s research also supports a shift in how postoperative care is delivered: moving from hospital-based observation to home-based continuous monitoring. “Patients go home, but now we can keep an eye on them,” she says.
And while data collection remains a challenge, she is optimistic. She hopes to build a dynamic, self-updating model that adapts as new data comes in. “That is the future: personalised, evolving risk prediction based on longitudinal data,” she says. In addition to her research, Fatime is committed to contributing to academic development through teaching and supervision. Soon, she plans to guide students and contribute to courses in biomedical engineering and artificial intelligence.
With two and a half years remaining in her PhD, her vision remains focused: to make healthcare smarter, earlier, and more proactive. “We cannot eliminate all complications,” she concludes, “but we can make sure they do not take us by surprise.”
