• Integrating Machine Learning with Magnetic Nanoparticle–Based Hyperthermia for Precision Cancer Therapy: A Targeted Review of Recent Evidence
  • Robab Bahreini,1,* Saman Alihosseini,2 Zohre Farrar,3
    1. Student Research and Technology Committee, Bushehr University of Medical Sciences, Bushehr, Iran
    2. Student Research and Technology Committee, Bushehr University of Medical Sciences, Bushehr, Iran
    3. Student Research and Technology Committee, Bushehr University of Medical Sciences, Bushehr, Iran


  • Introduction: Magnetic nanoparticle (MNP)–mediated hyperthermia has been recognized since the early 2000s as a primary or adjunctive modality for targeted cancer therapy. This treatment induces tumor cytotoxicity through localized heating; however, clinical application remains limited by variability in heating efficiency (specific loss power, SLP), heterogeneous intratumoral distribution, and the lack of real-time control. To achieve precision therapy and optimal therapeutic outcomes, integration of machine-learning (ML)–based predictive and control tools appears both beneficial and necessary.
  • Methods: A targeted review was conducted of studies published between 2015 and 2025 in which ML or deep-learning approaches were combined with MNP design, SLP prediction, imaging-guided thermometry, or closed-loop alternating magnetic field (AMF) control. PubMed, Scopus, and Web of Science were searched using the keywords: (Magnetic nanoparticles [Title/Abstract]) AND (Hyperthermia [Title/Abstract]) AND (Machine learning [Title/Abstract]) AND ((Oncology [Title/Abstract]) OR (Cancer [Title/Abstract])).Inclusion criteria required either experimental validation or simulation directly linking ML outputs to measurable hyperthermia parameters
  • Results: A total of **19 studies** initially met the inclusion criteria. After full-text screening and quality assessment. Across these, supervised ML models—including Random Forest, Gradient Boosting, and Deep Neural Networks—accurately predicted SLP and magnetic properties from physicochemical descriptors (size, composition, surface chemistry), reducing the need for extensive empirical testing. ML-enhanced imaging pipelines improved MNP localization and MR-thermometry reconstruction, enabling more precise spatiotemporal temperature estimation. Early closed-loop controllers using ML surrogate models demonstrated closer adherence to target temperature profiles in both simulations and preclinical settings.
  • Conclusion: ML integration across the MNP-hyperthermia workflow—ranging from nanoparticle synthesis optimization to in-vivo monitoring and adaptive AMF control—substantially improves predictive accuracy and thermal precision. Standardized datasets, physics-informed algorithms, and prospective preclinical validation will be critical next steps to accelerate clinical translation.
  • Keywords: magnetic nanoparticles; hyperthermia; machine learning; specific loss power; cancer therapy