Dr. ir. Martijn Starmans is Assistant Professor at the Erasmus MC (Rotterdam, the Netherlands) leading the AI for Integrated Diagnostics (AIID) research line, with a dual appointment at the Dept. of Radiology & Nuclear Medicine (BIGR group) and Dept. of Pathology (PHANTOM group). His vision is that we can and should learn more from previous studies, which he pursues by working on meta-level methods across clinical applications. His main research interest is the use of AI to improve the diagnostic work-up in oncology through integrated diagnostics, focussed on radiology (“radiomics”) and pathology (“pathomics”). Specifically, his group develops multi-modal machine learning methods to simultanously co-learn from both modalities (“radiopathomics”), and automated machine learning and meta-learning methods to generalize these methods across clinial applications. To establish his research line, Martijn received a prestigious 2M Euro NGF AINed Personal Fellowship in 2024. Martijn actively collaborates with of the BCN-AIM lab of prof. Lekadir at the University of Barcelona, with whom he created the FUTURE-AI guideline for trustworthy AI. He works on a variety of clinical applications (e.g. sarcoma, liver cancer, colorectal cancer, bladder cancer, melanoma, neuroendocrine tumors).
Collaborations & Consortia Martijn was part of the MICCAI 2024 organization committee one of the two first ever Open Data chairs. To promote sharing of medical imaging data, he organized the first ever Open Data Event at MICCAI 2024, with a focus on underrepresented diseases and populations, with for this year special attention for African datasets. To facilitate sharing of data, he established the AFRICAI repository, which other resources can use to make data FAIR.
Martijn is one of the initiators and PIs of the Sarcoma Artificial Intelligence (SAI) consortium (grant awarded), the Liver AI (LAI) consortium (grant awarded), and project lead of the Colorectal Liver Metastes AI (COLIMA) consortium (grant submitted). In these consortia, in total 51 clinical centers, companies, professional- and patient associations from 18 countries are united. He is also part of the Trustworthy AI for MRI ICAI LAB.
Martijn has a strong affinity with research infrastructure, which besides the MICCAI Open Data initiatives he pursues in various large European Horizon projects. He is a co-applicant of AFRICAI-RI: A Pan-African Research Infrastructure for Collaborative Biomedical Imaging and Artificial Intelligence in Respiratory Care, is work package leader of EuCanImage: Towards a European cancer imaging platform for enhanced Artificial Intelligence in oncology, and EOSC4Cancer: A European-wide foundation to accelerate Data-driven Cancer Research. Additionally, he is external advisor of RadioVal, and member of EUCAIM.
For a better overview on his projects, see his BIGR home page: https://bigr.nl/member/martijn/.
PhD Degree
Martijn obtained his PhD degree ‘‘cum laude’’ on February 1 2022 with his thesis titled Streamlined Quantitative Imaging Biomarker Development: Generalization of radiomics through automated machine learning. Following his passion to efficiently and automatically optimize routines, he developed an adaptive radiomics framework using automated machine learning, described in this paper. He collaborated with a large number of clinicians to develop radiomics biomarkers in a wide variety of clinical applications. His thesis was nominated for the Fredrik Philipsprijs 2023 for Best Dutch Radiology Thesis and made the top five.
Download my PhD Thesis.
Educational activities
Martijn has co-founded two courses and was teaching three courses in total in the BSc and MSc Technical Medicine. He is currently teaching one course in the MSc Applied Physics of the TU Delft. Martijn also was part of the program committee and mentor in the first AFRICAI Summer School, and part of the first Rotterdam Radiology AI course organized by the Dutch Radiology Society. He enjoys working together with students and has so-far (co-)supervised 53 students in interships and thesis projects.
Download my (slightly outdated) resumé .
PhD ''Streamlined Quantitative Imaging Biomarker Development'' ("cum laude"), 2022
Erasmus Medical Center, Rotterdam, the Netherlands
MSc Applied Physics, 2016
Delft University of Technology, Delft, the Netherlands
BSc Applied Physics, 2013
Delft University of Technology, Delft, the Netherlands
*Not officially mentioned as Co-PI due to formalities. Reference of formal applicant available upon request.
*Not officially mentioned as Co-PI due to formalities. Reference of formal applicant available upon request.
Despite major advances in artificial intelligence (AI) research for healthcare, the deployment and adoption of AI technologies remain limited in clinical practice. This paper describes the FUTURE-AI framework, which provides guidance for the development and deployment of trustworthy AI tools in healthcare.
Very proud of the first paper of PhD candidate Erik Kemper! This paper advocates for value-based development of AI tools in radiology, emphasizing that their limited clinical adoption stems from issues of usefulness and trust. By incorporating early health technology assessments (HTA) into the AI tool lifecycle, this study shows how a systematic evaluation can enhance clinical relevance and facilitate adoption into practice.
Very proud of the first paper of PhD candidate Douwe Spaanderman! This study externally and prospectively validated a radiomics model for differentiating lipomas from atypical lipomatous tumors (ALTs) on MRI, incorporating automatic and minimally interactive segmentation methods to improve clinical feasibility. The model demonstrated strong diagnostic performance across three large cohorts, achieving AUCs of up to 0.89 and performing comparably to expert radiologists, potentially reducing the need for biopsies.
Liver cancer is one the most common cancers worldwide. Magnetic resonance imaging (MRI) plays an important role in its diagnosis, but due to the wide variety of malignant and benign liver lesions, accurate diagnosis can be extremely challenging and highly reader dependent. The Liver Artificial Intelligence (LAI) consortium will develop novel MRI analysis methods to support the diagnosis of liver lesions and thereby improve treatment decisions for the individual patient.
Very proud of the first paper of PhD candidate Xinyi Wan! This study aims to improve the preoperative classification of nerve sheath tumors using radiomics, a method that extracts quantitative data from medical images. By analyzing MRI scans, we seek to develop a more accurate way to distinguish between different types of nerve sheath tumors before surgery. Our findings could lead to better treatment planning and outcomes for patients with these tumors. This research has the potential to enhance the diagnostic process and contribute to more personalized care for individuals with nerve sheath tumors, ultimately benefiting the medical community and patients alike.
Interview for the Beeldspraak magazine of the department of Radiology and Nuclear Medicine of the Erasmus MC on my new AIID research line.
The aim of the AI for Integrated Diagnostics (AIID) research line is to join forces of radiomics and pathomics to create trustworthy models to aid clinicians in decision making. Read more on the project page or the news article on the grant.
Radiomics uses quantitative medical imaging features and AI to create predictive models which can be used as biomarkers. In this thesis, we have developped an adaptive radiomics framework to automatically optimize the radiomics workflow per application and demonstrate its use to create biomarkers in eight different clinical applications.
We work on various clinical applications with different clinicians, and various technical solutions within medical image analysis & AI. The projects range from very technical hybrid, to more clinical. There are usually too many projects to describe here. Currently, we do not have any predefined projects available, we expect these to soon be put online, but we always have more than enough ideas, so feel free to contact me for more information if you are interested in an internship / thesis! For an overview of some past projects, please see https://bigr.nl/open-projects/.
*With a joint appointment at the Radiology & Nuclear and Pathology departments, Martijn is heading the AI for Integrated Diagnostics (AIID) research line. The AIID group develops novel multimodal machine learning methods to develop quantitative biomarkers, focused on medical imaging and application in concology.
Extending the work of his PhD, Martijn’s research focused on generalization of radiomics biomarkers (conventional and deep learning based) using automated machine learning and meta-learning. He worked on a variety of clinical applications (e.g. sarcoma, liver cancer, colorectal cancer, bladder cancer, melanoma, cardiomyopathy, neuroendocrine tumors, CRPS).
Martijn will visit the group of Prof. Dr. Karim Lekadir to collaborate on AI for oncology and cardiac imaging within the context of the EuCanImage and euCanSHare consortia. While the main focus is on developing novel algorithms for (deep) radiomics, they will also work on federated learning, applying Martijn’s adaptive radiomics framework from his PhD to ``new’’ clinical applications, a second version of the FUTURE-AI guiding principles, and improving the clinical utility of (radiomics) AI tools.
Supervisors: Karim Lekadir
See extended description in the “about” section on this website.
Supervisors: Stefan Klein, Jan-Jaap Visser, Wiro Niessen