Can AI Predict Treatment Response from Baseline Imaging?

Author: Rasit Dinc

The Dawn of a New Era in Personalized Medicine

In the realm of oncology, the ability to predict how a patient will respond to treatment is the cornerstone of personalized medicine. Foreseeing the efficacy of a therapeutic intervention before its administration can dramatically enhance clinical outcomes, mitigate the toxic effects of futile treatments, and optimize the allocation of healthcare resources. For decades, clinicians have relied on a combination of clinical assessments, histopathological findings, and molecular markers to forecast treatment response. However, these conventional methods are not without their limitations. They can be invasive, prone to subjective interpretation, and often fail to encapsulate the intricate heterogeneity of a tumor. [1]

Baseline medical imaging, including modalities such as magnetic resonance imaging (MRI), computed tomography (CT), and positron emission tomography (PET), offers a non-invasive glimpse into the tumor's microenvironment. The advent of artificial intelligence (AI), particularly in the domains of machine learning and deep learning, has ignited a surge of interest in harnessing the wealth of data embedded within these images to predict treatment response with unprecedented accuracy. [2]

The Synergy of AI and Radiomics

Radiomics, a burgeoning field in medical imaging, focuses on the high-throughput extraction of quantitative features from medical images. These features, often too subtle for the human eye to discern, can provide a granular characterization of the tumor phenotype. When coupled with sophisticated AI algorithms, radiomics can be instrumental in constructing predictive models for a spectrum of clinical endpoints, including the all-important treatment response.

A pivotal study recently published in Scientific Reports delved into the application of an automated radiomics model for predicting therapy response in multiple myeloma using baseline MRI scans. The research revealed that a model predicated solely on radiomics features exhibited a remarkable predictive accuracy, underscoring the potential of this non-invasive methodology in forecasting treatment response. The model, which was trained on a multi-center dataset, demonstrated its robustness and generalizability, two critical factors for clinical adoption. [1]

Similarly, a comprehensive review in Frontiers in Oncology underscored the transformative role of AI-driven image classification in predicting treatment response and prognosis in breast cancer. By meticulously analyzing pre-treatment imaging and pathology data, AI algorithms can effectively stratify patients into cohorts of likely responders and non-responders. This stratification empowers clinicians to customize treatment regimens, thereby providing a more robust and individualized risk assessment. [2]

Navigating the Challenges and Charting the Future

Despite the promising horizons, the widespread clinical implementation of AI-based treatment response prediction is not without its hurdles. The development of robust and reliable AI models necessitates large, diverse, and multi-center datasets for both training and validation. Furthermore, the lack of standardization in image acquisition protocols and analysis techniques across different institutions poses a significant challenge to the generalizability of these models. The interpretability of AI models, often referred to as the "black box" problem, is another critical aspect that needs to be addressed to foster trust and understanding among clinicians.

Future research endeavors will likely concentrate on the development of more sophisticated and generalizable AI models. The integration of multi-modal data, encompassing imaging, genomics, proteomics, and clinical data, is expected to yield more comprehensive and accurate predictive models. Moreover, the execution of prospective clinical trials is imperative to rigorously validate the clinical utility and cost-effectiveness of these AI-powered predictive tools.

Conclusion: A Paradigm Shift in Cancer Care

In conclusion, the application of AI to the analysis of baseline imaging represents a paradigm shift in the prediction of treatment response in oncology. By offering a non-invasive, quantitative, and holistic assessment of the tumor, the synergy of AI and radiomics has the potential to revolutionize the landscape of personalized medicine and significantly improve patient outcomes. While the path to routine clinical integration is paved with challenges, the relentless pace of innovation in AI and medical imaging promises a future where treatment decisions are more precise, personalized, and effective than ever before.

References

[1] Bauer, F., Hajiyianni, M., Weinhold, N. et al. Automated radiomics model for prediction of therapy response and minimal residual disease from baseline MRI in multiple myeloma. Sci Rep 15, 35400 (2025). https://doi.org/10.1038/s41598-025-13165-2

[2] Li, W., Chen, S., & Li, S. (2025). Predicting breast cancer treatment response and prognosis using AI-based image classification. Frontiers in Oncology, 15, 1619994. https://doi.org/10.3389/fonc.2025.1619994