Oral Presentations

Jonas Bohn

German Cancer Research Center

RPTK: The Role of Feature Computation on Prediction Performance

A rise of radiomics studies and techniques could be observed over the past few years, which centers around the extraction and analysis of quantitative features from medical images. Radiomics offers numerous advantages in disease characterization and treatment response prediction. Despite its promise, radiomics faces challenges in standardizing features and techniques, leading to large variations of approaches across studies and centers, making it difficult to determine the most suitable techniques for any given clinical scenario. Additionally, manually constructing optimized radiomics pipelines can be time-consuming. Recent works (WORC, Autoradiomics) have addressed the aforementioned shortcomings by introducing radiomics-based frameworks for automated pipeline optimization. Both approaches comprehensively span the entire radiomics workflow, enabling consistent, comprehensive, and reproducible radiomics analyses. In contrast, finding the ideal solutions for the workflow’s feature extractor and feature selection components, has received less attention. To address this, we propose the Radiomics Processing Toolkit (RPTK), which adds comprehensive feature extraction and selection components from PyRadiomics and from the Medical Image Radiomics Processor (MIRP) to the radiomics automation pipeline. To validate our approach and demonstrate benefits from the feature-centered components, we comprehensively compared RPTK with results from WORC and Autoradiomics on six public benchmark data sets. We show that we can achieve higher performance by incorporating the proposed feature processing and selection techniques. Our results provide additional guidance in selecting suitable components for optimized radiomics analyses in clinical use cases such as treatment response prediction.

Keywords: Radiomics *** Evaluation *** Automated Radiomics Processing *** Radiomics Processing Toolkit

You can find a pre-print version of this contribution here!

Oriane Thiery

Nantes Université, Centrale Nantes, CNRS, LS2N

Graph-based multimodal multi-lesion DLBCL treatment response prediction from PET images

Diffuse Large B-cell Lymphoma (DLBCL) is a lymphatic cancer involving one or more lymph nodes and extranodal sites. Its diagnostic and follow-up rely on Positron Emission Tomography (PET) and Computed Tomography (CT). After diagnosis, the number of non-responding patients to standard front-line therapy remains significant (30-40%). This work aims to develop a computer-aided approach to identify high-risk patients requiring adapted treatment by efficiently exploiting all the information available for each patient, including both clinical and image data. We propose a method based on recent graph neural networks that combine imaging information from multiple lesions, and a cross-attention module to integrate different data modalities efficiently. The model is trained and evaluated on a private prospective multicentric dataset of 583 patients. Experimental results show that our proposed method outperforms classical supervised methods based on either clinical, imaging or both clinical and imaging data for the 2-year progression-free survival (PFS) classification accuracy.

Keywords: Multimodal data fusion *** Graph Neural Networks *** Cross-attention *** DLBCL *** Treatment Response *** PET