Stefan Klein

Erasmus MC

In short, Stefan Klein's research entails the following: 

'Automatic grading and phenotyping of soft-tissue tumors through machine learning to guide personalized cancer treatment

Soft tissue tumors (STT) are a rare and complex group of lesions with a broad range of differentiation. All STT subtypes greatly differ in their clinical behavior, aggressiveness, molecular background, and preferred treatments given. Diagnosis of the correct phenotype, the grade of aggressiveness, and molecular make-up is therefore of utmost importance. Diagnosis of STT is generally supported by imaging, such as computed tomography (CT) and magnetic resonance imaging (MRI). However, visual assessment by a radiologist tends to be subjective and not precise. Quantitative, computational (“radiomics”) imaging features and state-of-the-art Artificial Intelligence (AI) techniques based on machine learning could enable more objective and precise STT diagnosis. With the support of the Hanarth Foundation, we aim to develop a comprehensive STT diagnostic model, both for phenotyping and grading. This model will be trained and validated in a large, multi-center cohort, and evaluated in a clinical setting. The model will be based on quantitative image analysis by radiomics and deep learning. We hypothesize that, by considering multiple STT phenotypes at once instead of training a specialized model for each subtype, breakthroughs will be achieved with regard to the diagnostic performance of the AI model and its generalizability. Our AI model will guide diagnosis and treatment decisions, thereby facilitating personalized medicine.'

Short summary of progress / results February 2024

This year marked a significant advancement in our research, as we successfully validated an AI method leveraging quantitative imaging features (radiomics) to differentiate different types of fatty tumors. We tested this method on 444 patients from hospitals in the United States, United Kingdom, and Netherlands, and it worked very well. Moreover, we enhanced our findings by integrating automatic and interactive segmentation methods developed previously. This means it's not only accurate but also practical for realworld clinical settings, providing a significant step towards clinical application of our non-invasive decision model for patients with fatty tumors.

Furthermore, we explored the broader utility of our segmentation methods in several downstream applications. Firstly, we used our segmentation method in combination with radiomics models to distinguish between malignant peripheral nerve sheath tumors (MPNST) and non-MPNST conditions (diseases in differential diagnosis). Secondly, we utilized these methods to accurately calculate total tumor volume to monitor treatment effectiveness in patients with high-grade sarcomas. Additionally, we demonstrated the versatility of our approach by effortlessly segmenting other tumor types 'out of the box,' eliminating the need for method reconfiguration or retraining.

In parallel, our team welcomed two new PhD candidates, Xinyi Wan and Matthew Marzetti, bolstering our efforts in sarcoma AI research. Together, we initiated a systematic review of existing AI methodologies for sarcoma, benchmarking them against best practices and evaluating their clinical applicability. Our aim is to ensure these methods adhere to the highest standards before integration into clinical practice, while also identifying potential areas for further focus and improvement

Brief summary of progress / results March 2023

In previous research we have developed artificial intelligence (AI) methods for soft tissue tumor classification based on quantitative imaging features (radiomics). However, these AI methods require a 3D delineation of the tumor, known as segmentation. This segmentation currently has to be made by a clinician, which means that these methods are impractical in clinic. For that reason, we focused on developing an interactive segmentation method in which the clinician only draws a few points in the tumor. Using this method we can make accurate segmentations for all types of soft tissue tumors. This method and previous radiomics methods are currently validated in several studies aimed at achieving accurate diagnosis of soft tissue tumors based on radiological imaging in clinical practice.

Brief summary of progress / results March 2022

Within this project we aim to develop an artificial intelligence (AI) model supporting soft-tissue tumor diagnosis, by predicting tumor phenotype and grade based on MRI and/or CT imaging. In the first year of this project, we have focused on:

1. development of a roadmap with clinically relevant diagnostic questions that might be tackled with AI;
2. data collection;
3. development of automated methods for tumor segmentation.

Regarding the last point, segmentation is prerequisite of many follow-up analyses by artificial intelligence methods. However, manual segmentation is a time-consuming process. We have therefore investigated the use of minimally interactive segmentation methods. One in which the clinician only is required to annotate a few points within the tumor, as well as a self-supervised learning approach in which the training of the segmentation model is initialized by a model trained to solve an auxiliary task for which more data is available.