Redona Brahimetaj is a Post-Doctoral researcher at the Department of Electronics and Informatics (ETRO) at the Vrije Universiteit Brussel. Driven by curiosity, her research spans multiple areas, with a main focus on: (a) developing Computer-Aided Detection and Diagnosis (CAD) to early detect breast cancer based only on high resolution 3D microcalcifcation images; (b) advancing gait analysis algorithms for both normal and pathological gait, using sensors like Inertial Measurement Units (IMUs) as a practical alternative to traditional motion capture systems like VICON. At present, she is involved in the European-funded project “Digilung” developing deep learning algorithms to analyse data (chronic obstructive pulmonary disease patients) from sensors embedded in a smart shirt.
Classification of tumor types: analysing microcalcification properties in 3D high resolution micro-CT images
Breast cancer, one of the most common cancers in women, often manifests early as microcalcifications (MCs) – tiny calcium deposits in breast tissue. Differentiating between benign and malignant MCs is challenging with standard 2D mammography, which suffers from limited contrast and resolution. Since the most realistic way to characterise a 3D object is to use a high-resolution 3D imaging modality, Redona focused on her PhD to analyse MCs which are visualised in high-resolution 3D. As in vivo imaging is still not possible, the images used are obtained from biopsies and scanned using a micro-CT scanner.
Her PhD contributions include: (1) developed a Computer-Aided Detection (CAD) system for breast cancer diagnosis using only features from individual MCs; (2) evaluated the impact of imaging resolution (8μm, 16μm, 32μm, 64μm) on MC classification; (3) conducted sensitivity analysis on segmentation thresholds and contributed to data collection, and (4) assessed deep learning frameworks for benign/malignant classification using 3D MC images.
Although the research is currently not directly applicable in vivo (3D high resolution in vivo breast imaging is still not possible), the work demonstrated its potential to be used in further research/clinical scenarios as soon as further improvements in the current breast screening modalities will occur. At the moment, the results achieved can potentially be used in intra operative imaging to reduce the waiting time between tissue extraction and anatomopathological results. As a long term goal, with our study we aim to avoid unnecessary biopsies and considerably reduce costs for the healthcare system.
Gait analysis using IMU sensors
Another key research focus is advancing gait analysis using IMUs. Unlike traditional motion capture systems such as VICON, which are costly, complex, and impractical for widespread use – particularly in pathological populations – IMUs provide a portable and cost-effective alternative for analysing gait and spatiotemporal gait parameters. Despite extensive research using motion capture systems as the gold standard, there remains a critical need to validate IMUs in accurately detecting three essential gait events: initial contact, final contact, and side detection (left/right). Errors in detecting these events can significantly impact spatiotemporal parameter calculations and the overall reliability of gait analysis. Our research aims to address these challenges by: (1) establishing a fundamental understanding and rigorous validation of IMU-based gait analysis for precise gait event detection; (2) developing deep learning algorithms to enhance the accuracy and reliability of gait event detection, and (3) bridging the gap between IMU-based methods and gold-standard motion capture systems to enable accurate and reliable gait analysis for both healthy and pathological populations.