Radiomics has been one of the most promising developments in medical imaging over the past decade. By extracting quantitative features from CT, MRI, or PET scans, radiomics promises to turn images into powerful biomarkers—guiding diagnosis, prognosis, and treatment response assessment. Yet, despite thousands of research papers, only a handful of radiomic signatures have made their way into routine clinical use.
So, where is the gap?
The Reproducibility Challenge
A major barrier is reproducibility and standardization. Radiomic features can vary depending on scanner type, acquisition protocols, reconstruction algorithms, and even the software used for feature extraction. This means a biomarker that looks reliable in one study may fail when tested across multiple centers or time points.
The METRICS-E³ initiative (radiomic.github.io/METRICS-E3) highlights this issue clearly. It emphasizes three essential pillars—Evaluation, Establishment, and Enablement—as prerequisites for trustworthy radiomics. Without standardized evaluation pipelines and transparent reporting, even the most exciting findings risk being irreproducible.
What the Evidence Says
Recent systematic reviews reinforce this concern.
- Currie et al. (2023) show that while radiomics has demonstrated impressive predictive power in oncology, lack of harmonization and insufficient validation remain the main obstacles to clinical translation (Insights Imaging, 2023).
- Lambin et al. (2025) provide updated evidence on the clinical value of radiomics, but again underline that reproducibility and rigorous prospective trials are the missing links between proof-of-concept and real-world impact (Insights Imaging, 2025).
Together, these works make it clear: radiomics is not failing due to lack of potential—it is failing due to variability and the absence of robust validation frameworks.
The Path Forward
To bridge this gap, the radiomics community needs to focus on:
- Calibration and harmonization across scanners and protocols.
- IBSI-compliant feature definitions to ensure comparability.
- Open, reproducible pipelines where methods can be independently verified.
- Prospective, multi-center validation studies that reflect clinical reality.
At CalibraMics, we believe this gap can be closed by combining standardized imaging phantoms with advanced drift detection and harmonization tools. Our mission is to help translate radiomics from promising research into trustworthy clinical tools that improve patient care.
References
Lambin, P. et al. “Clinical value of radiomics: an updated systematic review.” Insights Imaging, 16, 61 (2025). Link
METRICS-E³: Metrics for Evaluation, Establishment, and Enablement of Radiomics
Currie, G. et al. “Current evidence and future directions of radiomics in medical imaging.” Insights Imaging, 14, 23 (2023). Link