Frank Hoebers

Maastro

In short, Frank Hoebers' fellowship entails the following:

Predicting radiological extranodal extension in oropharyngeal carcinoma patients using AI

'In oropharyngeal squamous cell carcinoma, one of the most important prognostic factors is nodal status, including the presence of extranodal extension (ENE) with tumor cells infiltrating beyond the lymph node capsule into the surrounding tissues. In surgically treated patients, ENE can be diagnosed by pathological examination (pENE). However, most patients with oropharyngeal carcinoma are treated non-surgically by means of radiation or chemoradiation and thus information about pENE is lacking. Recently, extranodal extension based on radiological imaging (rENE) has also been associated with poor prognosis in these patients.

The scientific goal of this fellowship is to develop an AI tool that will support the radiologist in detecting rENE on radiological imaging.

Although AI has great potential for clinical practice, implementation in the clinic has proven to be difficult. This is in part attributed to the fact that for medical doctors it is often difficult to understand the underlying technology and computational methods that are used to create the algorithm. This may lead to lack of trust and skepticism. So, there is an urgent need to train clinicians in the field of AI.

The second goal of this fellowship is that I – as a clinician radiation-oncologist – will be trained to acquire knowledge and background skills in order to understand the development of AI software, with the particular use-case of detecting rENE on imaging. After this fellowship I will aim to become a future liaison between clinical data-scientists (developers) and the medical community (users) to increase support of the use of AI in daily routine.

Through generous funding by the Hanarth Fonds, I will be able to spend this personal fellowship at the Harvard Medical School in Boston, USA at the Artificial Intelligence in Medicine (AIM) Program.'

Brief summary of final results fellowship

August 2023. During my fellowship "Predicting radiological extranodal extension in oropharyngeal carcinoma patients using AI" in the past year, I have spent dedicated (research) time in order to learn about how to apply Artificial Intelligence in the field of Head and Neck Cancer. As a clinician (radiation- oncologist), it has been very valuable to work with experts in the field at the Artificial Intelligence in Medicine Program, Mass General Brigham, Harvard Medical School, Boston, Massachusetts with Prof. Hugo Aerts and dr. Benjamin Kann and colleagues and with the datascientists of my home-institution at the Clinical Data Science Group of Maastro Clinic, Maastricht.

My goal was to get a better understanding of the potential and applications of AI in the field of medicine from the perspective of a medical doctor. This was achieved by taking a series of theoretical courses and obtaining some basic hands-on programming skills, as well as by working in a multidisciplinary team of datascientists and doctors. The project focused on developing prediction models for prognosis in head and neck cancer patients using imaging (CT-scans) and applying deep learning techniques to predict the presence of extranodal extension in lymph node metastases in patients with oropharyngeal cancer. I also contributed to other projects in which autosegmentation tools were developed for primary tumors & lymph node metastases as well as for the delineation of muscle and fat tissue in the neck to assess the degree of sarcopenia (muscle wasting). These projects aim to better characterize the prognosis of patients and also try to find the optimal treatment approaches for the future.

With the knowledge obtained, I will now try to make a connection between the datascientists and the clinic in order to improve the implementation of AI models in daily practice. For this, I am planning to disseminate my expertise in AI to the medical community to increase trust of clinician-colleagues, resulting in improved understanding and acceptance of AI applications. By this process, I am expecting that the likelihood of implementation of AI tools will increase.