Postdoc or PhD developing artificial intelligence for accessible medical imaging

Medical imaging plays an essential role in disease detection, diagnosis, and aftercare. The accessibility to medical imaging is however challenged by the high costs of imaging devices, availability of the highly-trained personnel to operate these devices, and availability of experts to analyze and interpret the resulting images.

Our new public-private multi-institutional and multi-disciplinary consortium, "AI4AI: Artificial Intelligence for Accessible Medical Imaging", funded by Dutch Research Council, aims to address these bottlenecks. This will be achieved by developing artificial intelligence technologies to enable the use of more affordable devices, to allow operation by nonspecialized personnel, and to allow automated image interpretation.

We are seeking three highly motivated and talented PhD candidates or postdocs, where each focuses on one of the following applications:

1. Identification of patients at risk of coronary artery disease: CT is the primary modality for the detection of coronary artery disease in patients with stable chest pain. However, current CT imaging procedures are complex and require substantial expert time in image acquisition and interpretation. This hampers the use of cardiac CT imaging outside of hospitals or in resource-limited environments where affordable CT scanners are used. These scanners  typically generate images with suboptimal quality and therefore, make their automatic interpretation challenging. The researcher will develop AI-based algorithms to predict patient specific image acquisition parameters for optimal image quality and algorithms to autonomously quantify markers of coronary artery disease to extend the applicability of affordable scanners and reduce the need for specialized expert analysis.
2. Non-invasive identification of patients requiring invasive coronary artery treatment: Development of AI–based image analysis methods to identify patients requiring invasive coronary artery treatment using non-invasive coronary CT is an active field of research. However, these methods suffer from suboptimal image quality and heterogeneity in the image data. To extend the applicability of AI-based analysis, the researcher will develop methods that are robust to outliers and artefacts and enable identification of patients requiring invasive treatment at GPs, outpatient clinic settings or hospitals having access to CT. To enable autonomous employment of the developed methods, special attention will be paid to development of interpretable AI-based analysis.
3. Identification of fetal growth abnormality: The cornerstone of prenatal care is monitoring of fetal growth primarily to detect fetal growth restriction, a major cause of perinatal morbidity and mortality. Growth-restricted fetuses benefit from delivery before the onset of fetal hypoxia. Thus, timely identification of fetuses affected by growth restriction has the potential to reduce morbidity and mortality, especially in low- and middle-income countries. Obtaining accurate and complex measurement for reliable detection of fetal growth restriction requires a high level of expertise, limiting the analysis to specialized centers. Therefore, to allow high quality care, closer to patients’ home as well as in the areas where very specialized expertise is scarce, the researcher will develop autonomous AI-based systems that support fetal ultrasound acquisition and interpretation.

• You have a MSc degree in biomedical engineering, artificial intelligence, computer science, image sciences or a comparable training, with a strong interest in medical image processing and analysis (for the PhD position).
• Or you have a PhD degree in medical imaging (for the postdoc position).
• You have a good scientific background.
• You are highly motivated and can work well independently, but also in an interdisciplinary team of medical doctors and engineers.
• You have programming experience, preferably in Python. This is required.