Topic: Radiomics and Treatment Response Assessment in Cancer Therapy

Accurate and early assessment of treatment response is a cornerstone of precision oncology, guiding therapeutic decisions and informing prognostic predictions. Traditional approaches, such as those based on anatomical size changes (e.g., RECIST criteria), remain widely used in clinical practice. However, these methods offer limited insight into tumor biology and often detect treatment effects too late to meaningfully influence patient management. This underscores the pressing need for advanced, quantitative, and biologically informative biomarkers.

Radiomics - the high-throughput extraction of quantitative features from medical images - has emerged as a promising strategy to address this challenge. The field is now undergoing a paradigm shift fueled by artificial intelligence (AI). Beyond conventional hand-crafted radiomic features, deep learning and large foundational models (e.g., Vision Transformers, multi-modal models integrating imaging with genomic and clinical data) are unlocking new possibilities. These approaches can automatically discover complex, sub-visual patterns in medical images and seamlessly integrate them with multi-dimensional datasets, including clinical, pathological, and therapeutic information. This Special Issue aims to compile cutting-edge research and comprehensive reviews that highlight the transformative potential of AI-powered radiomics in redefining treatment response assessment. Topics of interest include novel methodologies such as self-supervised learning for small datasets, strategies to improve model interpretability, and approaches for multi-modal data fusion. Particular emphasis will be placed on clinical validation of these approaches for predicting and monitoring response across diverse cancer therapies, including immunotherapy, targeted therapy, radiotherapy, and chemotherapy. Ultimately, the goal is to accelerate the translation of AI-powered radiomics into routine clinical practice, foster personalized treatment strategies, improve patient outcomes, and bridge the gap between research innovation and bedside application.

Potential Topics:

AI and Methodological Innovations:

  ● Foundational models (e.g., Vision Transformers) for radionic feature extraction and analysis;

  ● Multi-modal large models integratingimaging, pathology, genomics, and clinical data;

  ● Deep learning for automated segmentation and response prediction;

  ● Self-supervised and transfer learning to address datascarcity;

  ● Explainable AI (XAI) for model interpretability and clinical adoption.

Clinical Applications in Specific Cancers:

  ● Predicting neoadjuvant therapy response in breast, rectal, and esophageal cancers;

  ● AI-based assessment of immunotherapy response in lung cancer and melanoma;

  ● Early prediction of radiotherapy efficacy in head and neck and brain tumors.

Multi-modal Data Integration:

  ● Radiogenomics: linking AI-derived imaging phenotypes with genomic biomarkers;

  ● Fusion of radiomics with clinical, pathological, and laboratory data.

Beyond Response: Forecasting Outcomes:

  ● AI-based models for predicting survival (PFS, OS) and recurrence;

  ● Imaging signatures for identifying adverse effects and treatment-related toxicity.

Translation and Challenges:

  ● Perspectives on clinical implementation of AI-powered radiomics;

  ● Addressing standardization, validation, and regulatory challenges.

Radiomics, treatment response assessment, artificial intelligence, foundation models, predictive biomarkers, cancer therapy, precision oncology