Open research questions in Radiomics and Machine Learning in Medical Imaging

What the literature leaves open

• The high dimensionality of the dataset. The need to ensure the robustness and reproducibility of the radiomics models. The challenge of selecting the most discriminative features from a large number of radiomics features.

  Predicting the biological behavior of cervical squamous cell carcinoma: a machine learning approach using apparent transverse relaxation rate (R2* maps) radiomics nomogram · 2026 · DOI

• The need for a more accurate prediction of biological behavior in cervical squamous cell carcinoma. The limitations of current methods in predicting deep stromal invasion, lymph node metastasis, and lymph-vascular space invasion. The potential of radiomics features to improve predictive performance.

  Predicting the biological behavior of cervical squamous cell carcinoma: a machine learning approach using apparent transverse relaxation rate (R2* maps) radiomics nomogram · 2026 · DOI

• Investigation of incremental predictive value of H&E-derived cellular composition beyond molecular testing, - Development of clinically meaningful explanations and accessible visualization tools for complex fusion architectures, - Prospective validation of multimodal fusion models

  Deep learning-based radiomics and pathomics for decoding tumor microenvironment and predicting immunotherapy outcomes in gastric cancer · 2026 · DOI

• The lack of clinically accessible approaches for spatially resolving the tumor immune contexture. Limited data availability and poor interpretability of deep learning-based radiomics and pathomics. Insufficient external validation of multimodal fusion models.

  Deep learning-based radiomics and pathomics for decoding tumor microenvironment and predicting immunotherapy outcomes in gastric cancer · 2026 · DOI

• Further research is needed to explore the use of inflammatory markers and clinical nomograms for irAE risk prediction, - The study suggests the need for proactive risk assessment in real-world HCC populations, - Future studies could investigate the application of the model in different cancer populations

  Development and external validation of an interpretable deep learning model for early immune-related adverse events in hepatocellular carcinoma · 2026 · DOI

• Lack of a reliable model for predicting early immune-related adverse events in patients with hepatocellular carcinoma. Need for a model that can identify patients at increased risk of clinically significant immune-related adverse events.

  Development and external validation of an interpretable deep learning model for early immune-related adverse events in hepatocellular carcinoma · 2026 · DOI

• Methodological heterogeneity and the need for prospective multicenter validation. The risk of overfitting and leakage in high-dimensional radiomics. The challenge of interpreting imaging signatures as non-invasive immune phenotypes.

  Radiomics and deep learning for immune checkpoint inhibitor outcomes: imaging-derived immune phenotyping and clinical translation across solid tumors · 2026 · DOI

• Limited dataset size, - Generalization across acquisition or institutional domains, - Feature selection and validation as methodological weaknesses, - Small sample size, - Methodological heterogeneity

  Radiomics and deep learning for immune checkpoint inhibitor outcomes: imaging-derived immune phenotyping and clinical translation across solid tumors · 2026 · DOI

• Although multimodal integration with MLEF (CB+CT+DP) was associated with higher performance, its incremental benefit remains uncertain, not translated in TEST and EXVAL.

  Clinical usability of an explainable AI decision support tool and evaluation of multimodal models in NSCLC · 2026 · DOI

• These results suggest that the prototype-guided adaptation can improve the use of large-scale unlabeled CT data for prognosis modeling when labeled survival data are limited.

  ProtoSurv: prototype-guided adaptation of computed tomography foundation model for lung-cancer prognosis prediction · 2026 · DOI

• Background: Therapeutic vulnerability in gastric cancer is profoundly influenced by the tumor microenvironment (TME), yet reliable and clinically actionable preoperative indicators remain insufficient.

  Therapeutic vulnerability shaped by the microenvironment: multi-omics and AI biomarkers for precision surgical planning in gastrointestinal tumors · 2026 · DOI

• Conclusions: Shifting from HRFs and SL to DRFs and SSL strategies, particularly in contexts with limited data points, enabling CT or PET alone, can significantly achieve high predictive performance.

  Enhanced Lung Cancer Survival Prediction Using Semi-Supervised Pseudo-Labeling and Learning from Diverse PET/CT Datasets · 2025 · DOI

• However, reliable biomarkers for identifying patients who are likely to benefit from NACI are lacking.

  Intratumoral microbiota-aided fusion radiomics model for predicting tumor response to neoadjuvant chemoimmunotherapy in triple-negative breast cancer · 2025 · DOI

• The improved stability at lower thresholds suggests that thoughtful annotation strategies can optimize the AI model training, particularly in contexts where training data are limited.

  The Three-Class Annotation Method Improves the AI Detection of Early-Stage Osteosarcoma on Plain Radiographs: A Novel Approach for Rare Cancer Diagnosis · 2024 · DOI

• Background/Objectives: Developing high-performance artificial intelligence (AI) models for rare diseases is challenging owing to limited data availability.

  The Three-Class Annotation Method Improves the AI Detection of Early-Stage Osteosarcoma on Plain Radiographs: A Novel Approach for Rare Cancer Diagnosis · 2024 · DOI

• Artificial intelligence (AI) can acquire characteristics that are not yet known to humans through extensive learning, enabling to handle large amounts of pathology image data.

  Artificial intelligence: illuminating the depths of the tumor microenvironment · 2024 · DOI

• The relationship between deep learning and biological signaling pathways in the context of concurrent chemoradiotherapy for non-small cell lung carcinoma has not been reported.

  Deep learning to estimate response of concurrent chemoradiotherapy in non-small-cell lung carcinoma · 2024 · DOI

• However, the use of deep learning (DL) models for predicting the response to CCRT in NSCLC remains unexplored.

  Deep learning to estimate response of concurrent chemoradiotherapy in non-small-cell lung carcinoma · 2024 · DOI

• Current diagnostic approaches for lung cancer face challenges such as radiation exposure, complex procedures, and high costs. There is a need for a non-invasive machine learning model for predicting the benign or malignant nature of subpleural lung lesions.

  Ultrasound-based deep learning radiomics for the differential diagnosis of benign and malignant subpleural pulmonary lesions · 2026 · DOI

• Accurate and timely identification of ALK rearrangements remains essential but can be limited by tissue availability, cost, and turnaround time of molecular testing. Heterogeneity in treatment response and the development of both primary and acquired resistance complicate long-term disease control. External validation and clinical implementation of AI models remain limited across studies.

  Artificial Intelligence in ALK-Rearranged NSCLC: Forecasting Response and Resistance · 2026 · DOI

• The study suggests that future research should focus on developing optimal optimization methods for deep learning models in medical imaging. The acquisition of large-scale medical image datasets for classification is a critical research priority. The study demonstrates the effectiveness of the multimodal model in preoperative breast cancer diagnosis, which can be further improved with future research.

  Multimodal Deep Learning Radiomics Nomogram for Preoperative Breast Cancer Prediction Using Ultrasound Imaging and Clinical Data · 2026 · DOI

• The acquisition of large-scale medical image datasets for classification is challenging because data are subject to ethical and privacy restrictions. The objective limitations that affect the performance of deep learning models in medical imaging create a critical research priority for developing optimal optimization methods. The study aims to address current research gaps and maximize the benefits of these technologies.

  Multimodal Deep Learning Radiomics Nomogram for Preoperative Breast Cancer Prediction Using Ultrasound Imaging and Clinical Data · 2026 · DOI

• Further research is needed to develop and validate AI-based methods for improving PET and SPECT imaging in AD. The use of deep learning and classical machine learning techniques should be explored further. The development of new radiotracers and imaging protocols should be continued.

  SPECT and PET imaging of Alzheimer’s disease revisited: from biomarkers to artificial intelligence-based prediction · 2026 · DOI

• Classical interpretation methods have limitations, including reader variability and dependence on reference regions. The use of AI-based methods can be limited by the quality of the input data and the complexity of the models. There is a need for standardization and harmonization to improve the comparability of results across different scanners and tracers.

  SPECT and PET imaging of Alzheimer’s disease revisited: from biomarkers to artificial intelligence-based prediction · 2026 · DOI

• The difficulty in diagnosing oral lesions due to subjective evaluations of clinical characteristics. The heterogeneity among the selected studies, which limited the ability to conduct formal quantitative syntheses. The limited availability of studies on the application of AI in oral cancer treatment.

  Artificial Intelligence in Oral Cancer: A Systematic Review · 2026 · DOI

Most-cited papers in Radiomics and Machine Learning in Medical Imaging

• Modern Diagnostic Imaging Technique Applications and Risk Factors in the Medical Field: A Review · BioMed Research International · 2022 · 698 citations
• The TRIPOD-LLM reporting guideline for studies using large language models · Nature Medicine · 2025 · 556 citations
• The Role of Artificial Intelligence in Early Cancer Diagnosis · Cancers · 2022 · 317 citations
• Sybil: A Validated Deep Learning Model to Predict Future Lung Cancer Risk From a Single Low-Dose Chest Computed Tomography · Journal of Clinical Oncology · 2023 · 268 citations
• METhodological RadiomICs Score (METRICS): a quality scoring tool for radiomics research endorsed by EuSoMII · Insights into Imaging · 2024 · 263 citations
• Development of a novel combined nomogram model integrating deep learning-pathomics, radiomics and immunoscore to predict postoperative outcome of colorectal cancer lung metastasis patients · Journal of Hematology & Oncology · 2022 · 258 citations
• Multistain deep learning for prediction of prognosis and therapy response in colorectal cancer · Nature Medicine · 2023 · 240 citations
• Predicting Microvascular Invasion in Hepatocellular Carcinoma Using CT-based Radiomics Model · Radiology · 2023 · 237 citations
• Convergence of evolving artificial intelligence and machine learning techniques in precision oncology · npj Digital Medicine · 2025 · 199 citations
• Foundation model for cancer imaging biomarkers · Nature Machine Intelligence · 2024 · 172 citations

Most recent work

• MRI-Based Radiomics for Non-Invasive Prediction of Molecular Biomarkers in Gliomas · Cancers · 2026
• Explainable Surgical Procedures Recommender System Leveraging Large Language Models · ACM Transactions on Recommender Systems · 2026
• Spatial radiomics-based interpretable multimodal machine learning model enhances outcomes prediction for minor stroke: A multicenter cohort study · Journal of Advanced Research · 2026
• A Novel CT Habitat Radiomics Approach for HER2 Status Prediction in Gastric Cancer · Academic Radiology · 2026
• PySERA: Open-source standardized python library for automated, scalable, and reproducible handcrafted and deep radiomics · Computer Methods and Programs in Biomedicine · 2026
• AI-enabled multi-omics integration in colorectal cancer: from molecular stratification to clinical translation · Frontiers in Cell and Developmental Biology · 2026
• Radiomics-integrated machine learning framework for quantitative breast cancer diagnosis · Frontiers in Artificial Intelligence · 2026
• Artificial Intelligence–Assisted Lung Nodule Evaluation on Low-Dose Chest CT in Asymptomatic Individuals: A Prospective Randomized Controlled Trial · American Journal of Roentgenology · 2026
• A new era of precision diagnosis and treatment for lung cancer: artificial intelligence-driven multimodal data integration and clinical applications · Cell Death & Disease · 2026
• Assessment of Robustness of MRI Radiomic Features in Four Abdominal Organs: Impact of Deep Learning Reconstruction and Segmentation · Journal of Magnetic Resonance Imaging · 2026

Related topics in Medicine