Discover how brain-computer interfaces connect neural signals, AI decoding, and digital twins in emerging neurotechnology.
Disclaimer: This report is intended for informational and educational purposes only and does not constitute medical advice, diagnosis, or treatment. The clinical protocols, AI-driven dosimetry models (e.g., 3D DosiNet), and combined-modality therapies discussed are based on current research and active clinical trials (such as DOORwaY90 and EMERALD-Y90) and may not yet be universally approved or standardized in all jurisdictions. Federal (USA) law restricts the sale and use of $^{90}\text{Y}$ microspheres and associated radiotherapy equipment to qualified physicians, and treatment decisions must be made in consultation with a multidisciplinary oncology team. Markedly abnormal synthetic or excretory liver function remains a contraindication for certain treatments described herein. Readers should consult with a licensed healthcare provider for specific medical guidance regarding liver cancer therapies, safety profiles, and associated risks.
Cancer care is entering a new phase as the line between internal radionuclide therapy and external beam radiation becomes increasingly fluid. At the center of this shift is Yttrium-90 (90Y), a powerful beta emitter long used in locoregional liver treatment, now being integrated with gamma-emitting isotopes and advanced external radiation techniques. Supported by rapid progress in artificial intelligence and a growing commitment to global health equity, this convergence is moving oncology beyond one-size-fits-all treatment toward a more precise and adaptive model of radiopharmaceutical medicine. By combining the targeted strength of 90Y with the imaging and therapeutic potential of gamma-based systems, next-generation radiation therapy is becoming more personalized, responsive, and globally accessible.
Current Clinical Reality of Y-90 and the Biophysical Landscape
Yttrium-90 transarterial radioembolization, also known as TARE or SIRT, has evolved from a late-stage rescue option into an important treatment consideration for unresectable hepatocellular carcinoma and liver-dominant metastatic colorectal cancer. Its effectiveness is closely tied to its physical properties. As a pure beta emitter, 90Y delivers highly localized radiation with a maximum energy of 2.27 MeV and a half-life of about 64 hours. Its average tissue penetration is roughly 2.5 mm, with a maximum range of up to 11 mm.
These characteristics make it especially well suited for targeted liver therapy. Delivered through the hepatic artery, 90Y can concentrate high tumoricidal doses within hypervascular tumors while largely sparing healthy liver tissue, which is supplied mainly through the portal vein. This selective distribution is one of the key reasons Y-90 has become such an important tool in modern locoregional oncology.
These characteristics make it especially well suited for targeted liver therapy. Delivered through the hepatic artery, 90Y can concentrate high tumoricidal doses within hypervascular tumors while largely sparing healthy liver tissue, which is supplied mainly through the portal vein. This selective distribution is one of the key reasons Y-90 has become such an important tool in modern locoregional oncology.
Microsphere Technologies: Glass vs. Resin
Yttrium-90 therapy is delivered through two main microsphere platforms: glass microspheres (TheraSphere) and resin microspheres (SIR-Spheres). Although both use the same isotope, their physical characteristics lead to different treatment dynamics and clinical strategies.
TheraSphere glass microspheres are smaller and carry a much higher specific activity per sphere. Because fewer spheres are needed to deliver the dose, they usually produce a lower embolic effect. SIR-Spheres, by contrast, are made from resin and involve a greater number of spheres per treatment, which increases the potential for arterial occlusion. In some tumors, especially those with particular vascular profiles, that difference may be clinically useful.
An important milestone in the development of resin microspheres came with the DOORwaY90 clinical trial. Results reported in July 2025 showed strong tumor control in treated patients, with no progression observed during one year of follow-up. The trial played a major role in supporting FDA approval of SIR-Spheres as a first-line treatment for unresectable hepatocellular carcinoma, marking an important shift in how radioembolization is positioned within liver cancer care.
At a glance: key differences
Resin microspheres (SIR-Spheres)
Biocompatible resin composition, approximately 50 Bq per sphere, diameter range of 20–60 μm, and a moderate-to-high embolic effect. FDA approved as a first-line option for HCC in 2025, with use in hepatocellular carcinoma and metastatic colorectal cancer.
Glass microspheres (TheraSphere)
Glass composition, approximately 2500 Bq per sphere, diameter range of 20–30 μm, and a low-to-moderate embolic effect. FDA approved for unresectable HCC in 2021, with use in hepatocellular carcinoma and cholangiocarcinoma.
Yttrium-90 therapy is delivered through two main microsphere platforms: glass microspheres (TheraSphere) and resin microspheres (SIR-Spheres). Although both use the same isotope, their physical characteristics lead to different treatment dynamics and clinical strategies.
TheraSphere glass microspheres are smaller and carry a much higher specific activity per sphere. Because fewer spheres are needed to deliver the dose, they usually produce a lower embolic effect. SIR-Spheres, by contrast, are made from resin and involve a greater number of spheres per treatment, which increases the potential for arterial occlusion. In some tumors, especially those with particular vascular profiles, that difference may be clinically useful.
An important milestone in the development of resin microspheres came with the DOORwaY90 clinical trial. Results reported in July 2025 showed strong tumor control in treated patients, with no progression observed during one year of follow-up. The trial played a major role in supporting FDA approval of SIR-Spheres as a first-line treatment for unresectable hepatocellular carcinoma, marking an important shift in how radioembolization is positioned within liver cancer care.
At a glance: key differences
Resin microspheres (SIR-Spheres)
Biocompatible resin composition, approximately 50 Bq per sphere, diameter range of 20–60 μm, and a moderate-to-high embolic effect. FDA approved as a first-line option for HCC in 2025, with use in hepatocellular carcinoma and metastatic colorectal cancer.
Glass microspheres (TheraSphere)
Glass composition, approximately 2500 Bq per sphere, diameter range of 20–30 μm, and a low-to-moderate embolic effect. FDA approved for unresectable HCC in 2021, with use in hepatocellular carcinoma and cholangiocarcinoma.
Guideline Evolution and Patient Selection
The growing role of 90Y^{90}\mathrm{Y}90Y in clinical guidelines, including those from the NCCN and BCLC, reflects increasing confidence in its safety and therapeutic value. Recent updates in 2025 and 2026 have placed greater emphasis on patient-specific dosimetry and the preservation of liver function, underscoring a more individualized approach to treatment planning.
Careful patient selection remains essential. In general, clinicians aim to keep total bilirubin below 2.0 mg/dL, with an even lower threshold of under 1.3 mg/dL for bilobar treatment, while the Child–Pugh score ideally should not exceed B7. Another critical step is assessment of the lung shunt fraction (LSF), since excessive shunting can expose the lungs to unintended radiation. An LSF above 20% is considered a major safety concern because it raises the risk of radiation pneumonitis, a potentially serious complication.
The growing role of 90Y^{90}\mathrm{Y}90Y in clinical guidelines, including those from the NCCN and BCLC, reflects increasing confidence in its safety and therapeutic value. Recent updates in 2025 and 2026 have placed greater emphasis on patient-specific dosimetry and the preservation of liver function, underscoring a more individualized approach to treatment planning.
Careful patient selection remains essential. In general, clinicians aim to keep total bilirubin below 2.0 mg/dL, with an even lower threshold of under 1.3 mg/dL for bilobar treatment, while the Child–Pugh score ideally should not exceed B7. Another critical step is assessment of the lung shunt fraction (LSF), since excessive shunting can expose the lungs to unintended radiation. An LSF above 20% is considered a major safety concern because it raises the risk of radiation pneumonitis, a potentially serious complication.
The Role of AI in Transforming Radiotheranostics
Artificial intelligence is becoming one of the key forces behind the shift toward precision medicine in radiotherapy. By helping clinicians manage complex dose patterns and reduce the computational burden of dosimetry, AI is improving decision-making across the entire treatment pathway—from pre-treatment planning and dose prediction to procedural guidance and post-treatment evaluation.
Generative AI and Dose Prediction
Among the most promising tools are Generative Adversarial Networks (GANs), which are being used for 3D dose prediction and activity map generation. In 990Y therapy, these models can translate pre-treatment 99mTc-MAA scans into predictions of post-treatment microsphere distribution. This offers clinicians a more detailed view of how treatment may behave before it is delivered.
Advanced voxel-based models, including variants of Pix2Pix, have shown that AI can estimate 90Y biodistribution with remarkable precision, including very small average dose differences in normal liver tissue. These systems can also generate synthetic CT or PET images that better reflect tissue heterogeneity, allowing clinicians to account for uneven uptake within a tumor and improve response prediction beyond what simplified voxel-based dosimetry can offer.
Deep Reinforcement Learning and Treatment Optimization
While generative models are powerful for prediction, Deep Reinforcement Learning (DRL) is opening new possibilities in treatment optimization. DRL systems can learn dosing strategies through repeated simulation in virtual treatment environments, making them especially relevant for combined-modality care.
This is particularly important when internal 90Y therapy is paired with external beam approaches such as SBRT, where the goal is to balance highly localized internal dose delivery with broader external coverage. One of the most promising developments is interactive planning, in which clinicians can adjust treatment trade-offs in real time using slider-based controls. This makes it possible to fine-tune the balance between tumor coverage and protection of organs at risk, creating plans that are both technically robust and more individualized.
Computational Dosimetry: From Monte Carlo to GPU Acceleration
Dosimetry remains one of the most computationally demanding parts of radiotheranostics. Traditional Monte Carlo simulations are still considered the gold standard for accuracy, but they have historically been too slow for widespread routine use. Newer approaches are changing that.
GPU-accelerated Monte Carlo methods and AI-based models such as 3D DosiNet are making high-accuracy dose mapping much faster and more practical. Built on deep learning architectures such as Res UNet, 3D DosiNet can generate absorbed dose maps that closely match Monte Carlo results while reducing processing time from hours to minutes—or even seconds in some workflows.
In practical terms, this creates a more flexible dosimetry ecosystem:
As these tools mature, they are helping reduce uncertainty in tumor and lung dose estimation, strengthening both safety and therapeutic precision. The broader significance is clear: AI is not replacing radiotheranostics, but making it more adaptive, more computationally efficient, and more responsive to the individual patient.
Generative AI and Dose Prediction
Among the most promising tools are Generative Adversarial Networks (GANs), which are being used for 3D dose prediction and activity map generation. In 990Y therapy, these models can translate pre-treatment 99mTc-MAA scans into predictions of post-treatment microsphere distribution. This offers clinicians a more detailed view of how treatment may behave before it is delivered.
Advanced voxel-based models, including variants of Pix2Pix, have shown that AI can estimate 90Y biodistribution with remarkable precision, including very small average dose differences in normal liver tissue. These systems can also generate synthetic CT or PET images that better reflect tissue heterogeneity, allowing clinicians to account for uneven uptake within a tumor and improve response prediction beyond what simplified voxel-based dosimetry can offer.
Deep Reinforcement Learning and Treatment Optimization
While generative models are powerful for prediction, Deep Reinforcement Learning (DRL) is opening new possibilities in treatment optimization. DRL systems can learn dosing strategies through repeated simulation in virtual treatment environments, making them especially relevant for combined-modality care.
This is particularly important when internal 90Y therapy is paired with external beam approaches such as SBRT, where the goal is to balance highly localized internal dose delivery with broader external coverage. One of the most promising developments is interactive planning, in which clinicians can adjust treatment trade-offs in real time using slider-based controls. This makes it possible to fine-tune the balance between tumor coverage and protection of organs at risk, creating plans that are both technically robust and more individualized.
Computational Dosimetry: From Monte Carlo to GPU Acceleration
Dosimetry remains one of the most computationally demanding parts of radiotheranostics. Traditional Monte Carlo simulations are still considered the gold standard for accuracy, but they have historically been too slow for widespread routine use. Newer approaches are changing that.
GPU-accelerated Monte Carlo methods and AI-based models such as 3D DosiNet are making high-accuracy dose mapping much faster and more practical. Built on deep learning architectures such as Res UNet, 3D DosiNet can generate absorbed dose maps that closely match Monte Carlo results while reducing processing time from hours to minutes—or even seconds in some workflows.
In practical terms, this creates a more flexible dosimetry ecosystem:
- MIRD schema offers fast organ-level estimates and remains useful for high-throughput settings.
- Voxel-based S-values improve spatial detail with moderate computational demand.
- Standard Monte Carlo provides the highest accuracy but is often too slow for routine high-volume practice.
- GPU-accelerated Monte Carlo preserves high accuracy while making clinical use far more realistic.
- AI models such as 3D DosiNet offer rapid, high-quality estimates with strong potential for scale.
As these tools mature, they are helping reduce uncertainty in tumor and lung dose estimation, strengthening both safety and therapeutic precision. The broader significance is clear: AI is not replacing radiotheranostics, but making it more adaptive, more computationally efficient, and more responsive to the individual patient.
Clinical Synergy: Integrating Y-90 with Gamma-Based Modalities
One of the most promising directions in modern radiotheranostics is the integration of internal 90Y therapy with external radiation techniques such as SBRT and EBRT. This combined approach helps address one of the main limitations of radionuclide therapy: the presence of cold spots, or regions of the tumor that receive less radiation because microsphere distribution depends on vascular flow. When tumor perfusion is uneven, some areas may be underdosed. External radiation can help compensate for this by delivering a more uniform boost to regions that internal therapy may not fully reach.
The Safety of Sequential Radiation
A key concern in combining these modalities is the risk of cumulative liver toxicity. Recent research, however, suggests that carefully sequenced treatment may be safer than previously assumed. Findings reported in 2026 from the University of Cincinnati Cancer Center indicated that EBRT administered after 90Y therapy did not increase the rate of radiation-induced liver disease or severe grade 3+ toxicities, even in patients who had already undergone TARE.
This has important clinical implications. It opens the possibility of more individualized sequencing, such as using 90Y first for debulking or downstaging, followed by focused SBRT to treat residual disease or help bridge a patient toward liver transplantation.
Theranostic Pairs and Molecular Precision
Another major advance comes from the use of theranostic isotope pairs, sometimes described as “theranostic twins.” These pair a diagnostic isotope used for imaging with a therapeutic isotope used for treatment, allowing clinicians to visualize targeting patterns before therapy begins.
The 86Y /90Y pair is a clear example: 86Y supports high-resolution PET imaging, helping map the biodistribution expected from 90Y treatment.
The field is also expanding beyond traditional targets such as SSTR and PSMA. New tracers aimed at fibroblast activation protein (FAP), including the 68Ga/177Lu-FAPT pair, are broadening the reach of radiotheranostics across many epithelial tumors. Because the same targeting ligand can be preserved while the isotope changes from diagnostic to therapeutic use, this model supports a more seamless transition from imaging to treatment and strengthens the logic of personalized dosimetry.
Synergy with Immunotherapy
The convergence of radiotherapy and immunotherapy is another area of growing interest. Both the beta particles emitted by 90Yand the radiation delivered through SBRT can reshape the tumor microenvironment, potentially making it more responsive to immune-based treatment. This possibility is now being explored in trials such as EMERALD-Y90, which combines 90Y glass microspheres with durvalumab and bevacizumab.
Early results from related studies suggest that TARE may do more than deliver radiation alone. By modulating the tumor microenvironment, it may also enhance immune response in selected patients, with potential gains in objective response rate and overall survival. While this area is still developing, it represents one of the most important frontiers in combined-modality liver cancer care.
The Safety of Sequential Radiation
A key concern in combining these modalities is the risk of cumulative liver toxicity. Recent research, however, suggests that carefully sequenced treatment may be safer than previously assumed. Findings reported in 2026 from the University of Cincinnati Cancer Center indicated that EBRT administered after 90Y therapy did not increase the rate of radiation-induced liver disease or severe grade 3+ toxicities, even in patients who had already undergone TARE.
This has important clinical implications. It opens the possibility of more individualized sequencing, such as using 90Y first for debulking or downstaging, followed by focused SBRT to treat residual disease or help bridge a patient toward liver transplantation.
Theranostic Pairs and Molecular Precision
Another major advance comes from the use of theranostic isotope pairs, sometimes described as “theranostic twins.” These pair a diagnostic isotope used for imaging with a therapeutic isotope used for treatment, allowing clinicians to visualize targeting patterns before therapy begins.
The 86Y /90Y pair is a clear example: 86Y supports high-resolution PET imaging, helping map the biodistribution expected from 90Y treatment.
The field is also expanding beyond traditional targets such as SSTR and PSMA. New tracers aimed at fibroblast activation protein (FAP), including the 68Ga/177Lu-FAPT pair, are broadening the reach of radiotheranostics across many epithelial tumors. Because the same targeting ligand can be preserved while the isotope changes from diagnostic to therapeutic use, this model supports a more seamless transition from imaging to treatment and strengthens the logic of personalized dosimetry.
Synergy with Immunotherapy
The convergence of radiotherapy and immunotherapy is another area of growing interest. Both the beta particles emitted by 90Yand the radiation delivered through SBRT can reshape the tumor microenvironment, potentially making it more responsive to immune-based treatment. This possibility is now being explored in trials such as EMERALD-Y90, which combines 90Y glass microspheres with durvalumab and bevacizumab.
Early results from related studies suggest that TARE may do more than deliver radiation alone. By modulating the tumor microenvironment, it may also enhance immune response in selected patients, with potential gains in objective response rate and overall survival. While this area is still developing, it represents one of the most important frontiers in combined-modality liver cancer care.
The Regulatory and Standardized Frontier
As AI and novel isotope combinations move more quickly into clinical practice, the need for validation and standardization has become increasingly urgent. International professional bodies are now working to close this gap, recognizing that advanced radiotherapy systems must be not only innovative, but also reproducible, transparent, and safe across institutions and national settings.
ESTRO–AAPM Guidelines for AI
A major step in this direction came in August 2024, when the European Society for Radiotherapy and Oncology (ESTRO) and the American Association of Physicists in Medicine (AAPM) released a joint guideline for the development and clinical validation of AI models in radiation therapy. Built through a Delphi process, the guideline establishes a shared framework for bringing AI into practice more responsibly.
Its priorities include:
Together, these 19 statements create a practical benchmark for high-quality AI in radiotherapy and help bridge the divide between technical innovation and clinical adoption.
Regulatory Engagement and FDA Guidance
Regulators are also beginning to respond more directly to the rise of AI-enabled treatment systems. In 2025, the U.S. Food and Drug Administration (FDA) released draft guidance on the use of AI in regulatory decision-making for drugs and biological products. This matters for radiotheranostics because AI-driven dosimetry software increasingly functions as part of the treatment infrastructure itself.
As these tools evolve toward Software as a Medical Device (SaMD), the expectation is clear: algorithms used to calculate or guide 90Y dosing must be held to standards as rigorous as those applied to the therapeutic agents and delivery systems they support. In this sense, regulation is no longer separate from innovation—it is becoming one of the conditions that makes safe innovation possible.
ESTRO–AAPM Guidelines for AI
A major step in this direction came in August 2024, when the European Society for Radiotherapy and Oncology (ESTRO) and the American Association of Physicists in Medicine (AAPM) released a joint guideline for the development and clinical validation of AI models in radiation therapy. Built through a Delphi process, the guideline establishes a shared framework for bringing AI into practice more responsibly.
Its priorities include:
- Transparent reporting, so that training, validation, and testing methods are clearly documented and AI systems do not function as opaque black boxes.
- Model availability, encouraging the sharing of models and datasets to support external verification and broader generalizability.
- Quality assurance, including both patient-specific QA and the ongoing monitoring and maintenance of AI software over time.
Together, these 19 statements create a practical benchmark for high-quality AI in radiotherapy and help bridge the divide between technical innovation and clinical adoption.
Regulatory Engagement and FDA Guidance
Regulators are also beginning to respond more directly to the rise of AI-enabled treatment systems. In 2025, the U.S. Food and Drug Administration (FDA) released draft guidance on the use of AI in regulatory decision-making for drugs and biological products. This matters for radiotheranostics because AI-driven dosimetry software increasingly functions as part of the treatment infrastructure itself.
As these tools evolve toward Software as a Medical Device (SaMD), the expectation is clear: algorithms used to calculate or guide 90Y dosing must be held to standards as rigorous as those applied to the therapeutic agents and delivery systems they support. In this sense, regulation is no longer separate from innovation—it is becoming one of the conditions that makes safe innovation possible.
Infrastructure, Global Access, and LMICs
One of the biggest challenges in modern radiotherapy is unequal access. While high-income countries often have broad treatment availability, many low-income countries still have access to only a small fraction of the radiotherapy capacity they need. As precision oncology advances, the question is no longer only how to improve treatment, but how to make it reachable.
The IAEA Rays of Hope Initiative
The IAEA’s Rays of Hope initiative has become one of the most important global efforts to address this imbalance. Launched in 2022, the program is focused on building sustainable radiotherapy systems in low- and middle-income countries.
Its work includes:
A major milestone came in July 2025, when Malawi opened its first public radiotherapy center at Kamuzu Central Hospital, offering treatment to patients who previously had little or no access to this level of care.
Decentralized Isotope Production: The Role of Accelerators
Another critical issue is isotope supply. Therapeutic isotopes such as 90Y^{90}\mathrm{Y}90Y still depend heavily on aging nuclear research reactors, which makes the supply chain vulnerable and unevenly distributed. For LMICs, long-term access will depend increasingly on more decentralized and scalable production methods.
Two of the most important alternatives are:
Small medical cyclotrons
These systems, typically operating below 20 MeV, are well suited for hospital-based production of PET isotopes and can strengthen local diagnostic capacity.
Compact linear accelerators
Linacs offer a more modular model with lower radioactive waste and a stronger safety profile, making them especially attractive for urban hospital settings and regional treatment hubs.
Canada’s Contribution
Canada is also contributing to this transition. Through a framework with the IAEA, the Canadian Nuclear Isotope Council (CNIC) is supporting training and technical development beginning in 2026. This includes programs at Bruce Power and major Canadian hospitals focused on isotope production, radiochemistry, and dosimetry. The broader goal is to help LMICs build the expertise needed for greater long-term self-sufficiency.
At a glance: production infrastructure options
Nuclear research reactors
Best suited for neutron-rich therapeutic isotopes, but they come with high waste production, complex safety requirements, very high maintenance costs, and a massive regional or national footprint.
Medical cyclotrons
Well suited for positron- and gamma-emitting diagnostic isotopes, with lower waste, easier supervision, moderate maintenance costs, and a hospital-scale footprint.
Compact linear accelerators
Increasingly attractive for targeted alpha- and beta-emitting therapeutic isotopes, with very low waste, high safety, lower maintenance burden, and a modular clinic-based design.
The IAEA Rays of Hope Initiative
The IAEA’s Rays of Hope initiative has become one of the most important global efforts to address this imbalance. Launched in 2022, the program is focused on building sustainable radiotherapy systems in low- and middle-income countries.
Its work includes:
- expanding access to essential diagnostic and treatment equipment, including linear accelerators and mammography units
- establishing Anchor Centres for training and professional development in radiation medicine
- supporting oncology workforce development across regions with limited infrastructure
A major milestone came in July 2025, when Malawi opened its first public radiotherapy center at Kamuzu Central Hospital, offering treatment to patients who previously had little or no access to this level of care.
Decentralized Isotope Production: The Role of Accelerators
Another critical issue is isotope supply. Therapeutic isotopes such as 90Y^{90}\mathrm{Y}90Y still depend heavily on aging nuclear research reactors, which makes the supply chain vulnerable and unevenly distributed. For LMICs, long-term access will depend increasingly on more decentralized and scalable production methods.
Two of the most important alternatives are:
Small medical cyclotrons
These systems, typically operating below 20 MeV, are well suited for hospital-based production of PET isotopes and can strengthen local diagnostic capacity.
Compact linear accelerators
Linacs offer a more modular model with lower radioactive waste and a stronger safety profile, making them especially attractive for urban hospital settings and regional treatment hubs.
Canada’s Contribution
Canada is also contributing to this transition. Through a framework with the IAEA, the Canadian Nuclear Isotope Council (CNIC) is supporting training and technical development beginning in 2026. This includes programs at Bruce Power and major Canadian hospitals focused on isotope production, radiochemistry, and dosimetry. The broader goal is to help LMICs build the expertise needed for greater long-term self-sufficiency.
At a glance: production infrastructure options
Nuclear research reactors
Best suited for neutron-rich therapeutic isotopes, but they come with high waste production, complex safety requirements, very high maintenance costs, and a massive regional or national footprint.
Medical cyclotrons
Well suited for positron- and gamma-emitting diagnostic isotopes, with lower waste, easier supervision, moderate maintenance costs, and a hospital-scale footprint.
Compact linear accelerators
Increasingly attractive for targeted alpha- and beta-emitting therapeutic isotopes, with very low waste, high safety, lower maintenance burden, and a modular clinic-based design.
Precision Radiopharmaceutical Medicine: The 2030 Vision
Looking toward the end of the decade, the integration of 90Y, gamma emitters, and AI-guided planning points toward a more connected model of radiation therapy. Rather than treating radiotherapy as a fixed sequence of isolated interventions, this vision imagines a more adaptive ecosystem—one that responds to the biology of the patient, the behavior of the tumor, and the evolving effects of treatment over time.
The Rise of the Digital Twin
One of the most compelling ideas in this future is the digital twin: a virtual model of a patient’s anatomy and disease biology that can be updated as new data becomes available. By combining serial imaging, biomarker data such as circulating tumor DNA, and other clinical inputs, digital twins may help support a more adaptive form of treatment planning. In this model, therapy could be adjusted in response to how a tumor is actually behaving. A strong response to 90Y radioembolization, for example, might support a lower follow-up SBRT dose, while resistant regions could receive a more focused boost.
Integrating Biology at the Micron Scale
Another major direction is the deeper integration of biology into treatment planning. Instead of relying only on anatomy and gross tumor volume, future systems may begin to map features such as hypoxia, heterogeneity, and DNA repair capacity at much finer resolution. AI-supported imaging analysis could help identify resistant subregions within a tumor, making it possible to deliver higher doses where they are most needed while sparing more vulnerable healthy tissue. This approach, often described as biological dose painting, represents a move toward radiotherapy that is shaped not only by where a tumor is, but by how it behaves.
Toward Long-Term Disease Control
The broader goal is not simply to intensify treatment, but to make cancer care more precise, repeatable, and sustainable over time. In this vision, highly targeted radionuclide therapy, optimized brachytherapy, FLASH radiotherapy, and smarter isotope-guided feedback systems may work together to support longer-term disease control with lower toxicity. Rather than a one-time intervention model, radiotherapy increasingly begins to resemble an adaptive management strategy—one designed to respond to change while preserving quality of life.
The Rise of the Digital Twin
One of the most compelling ideas in this future is the digital twin: a virtual model of a patient’s anatomy and disease biology that can be updated as new data becomes available. By combining serial imaging, biomarker data such as circulating tumor DNA, and other clinical inputs, digital twins may help support a more adaptive form of treatment planning. In this model, therapy could be adjusted in response to how a tumor is actually behaving. A strong response to 90Y radioembolization, for example, might support a lower follow-up SBRT dose, while resistant regions could receive a more focused boost.
Integrating Biology at the Micron Scale
Another major direction is the deeper integration of biology into treatment planning. Instead of relying only on anatomy and gross tumor volume, future systems may begin to map features such as hypoxia, heterogeneity, and DNA repair capacity at much finer resolution. AI-supported imaging analysis could help identify resistant subregions within a tumor, making it possible to deliver higher doses where they are most needed while sparing more vulnerable healthy tissue. This approach, often described as biological dose painting, represents a move toward radiotherapy that is shaped not only by where a tumor is, but by how it behaves.
Toward Long-Term Disease Control
The broader goal is not simply to intensify treatment, but to make cancer care more precise, repeatable, and sustainable over time. In this vision, highly targeted radionuclide therapy, optimized brachytherapy, FLASH radiotherapy, and smarter isotope-guided feedback systems may work together to support longer-term disease control with lower toxicity. Rather than a one-time intervention model, radiotherapy increasingly begins to resemble an adaptive management strategy—one designed to respond to change while preserving quality of life.
Synthesis and Strategic Recommendations
The future of integrating 90Y and gamma-emitting systems into next-generation radiation therapy will depend on the convergence of biophysics, computational intelligence, and global healthcare infrastructure. To move this field forward responsibly, several priorities stand out.
Expand multi-institutional data sharing
Robust AI models such as 3D DosiNet depend on large, diverse, and well-structured datasets. Expanding shared initiatives and standardized planning cohorts will be essential for improving model reliability, reducing data scarcity, and supporting broader clinical validation.
Validate combined-modality treatment in larger trials
Early evidence supporting sequential 90Y, SBRT, and related combined approaches is promising, but larger trials remain necessary. Phase III studies will be especially important for confirming safety, clarifying benefit, and defining where these strategies fit best—particularly when combined with immunotherapy.
Invest in decentralized isotope supply
Long-term progress will also depend on a more stable and equitable isotope infrastructure. Moving beyond heavy reliance on centralized reactor systems toward regional production supported by compact cyclotrons and linear accelerators could strengthen resilience, improve access, and support wider global use.
Strengthen global training standards
As radiotherapy becomes more data-driven, workforce development must evolve with it. International frameworks such as the ESTRO–AAPM guidelines and the IAEA Rays of Hope Anchor Centres offer an important foundation for training clinicians not only in isotope delivery, but also in the oversight, interpretation, and validation of AI-supported treatment systems.
By the early 2030s, the most mature radiotherapy pathways may be those that successfully combine molecular targeting, computational insight, and clinical expertise within a single adaptive model of care. The deeper goal is not only technological progress, but a more precise, equitable, and sustainable form of cancer treatment—one capable of delivering the right radiation to the right patient at the right time, across very different healthcare settings.
Expand multi-institutional data sharing
Robust AI models such as 3D DosiNet depend on large, diverse, and well-structured datasets. Expanding shared initiatives and standardized planning cohorts will be essential for improving model reliability, reducing data scarcity, and supporting broader clinical validation.
Validate combined-modality treatment in larger trials
Early evidence supporting sequential 90Y, SBRT, and related combined approaches is promising, but larger trials remain necessary. Phase III studies will be especially important for confirming safety, clarifying benefit, and defining where these strategies fit best—particularly when combined with immunotherapy.
Invest in decentralized isotope supply
Long-term progress will also depend on a more stable and equitable isotope infrastructure. Moving beyond heavy reliance on centralized reactor systems toward regional production supported by compact cyclotrons and linear accelerators could strengthen resilience, improve access, and support wider global use.
Strengthen global training standards
As radiotherapy becomes more data-driven, workforce development must evolve with it. International frameworks such as the ESTRO–AAPM guidelines and the IAEA Rays of Hope Anchor Centres offer an important foundation for training clinicians not only in isotope delivery, but also in the oversight, interpretation, and validation of AI-supported treatment systems.
By the early 2030s, the most mature radiotherapy pathways may be those that successfully combine molecular targeting, computational insight, and clinical expertise within a single adaptive model of care. The deeper goal is not only technological progress, but a more precise, equitable, and sustainable form of cancer treatment—one capable of delivering the right radiation to the right patient at the right time, across very different healthcare settings.
Recommended Reading List
1. Y-90 fundamentals and clinical practice
Start here for the clinical foundation of radioembolization, patient selection, and liver-directed therapy.
2. AI in radiotherapy and theranostics
Best for understanding how AI is reshaping planning, prediction, and treatment workflows.
3. Dosimetry and computational planning
For the technical bridge between classical planning and next-generation AI-supported dosimetry.
4. Combined modalities, theranostic pairs, and immunotherapy
Best for your section on synergy between Y-90, external radiation, and targeted molecular frameworks.
5. Regulation, validation, and standards
Important if you want the paper to sound serious and policy-aware.
6. Global infrastructure, access, and isotope production
For your LMIC and global access section.
1. Y-90 fundamentals and clinical practice
Start here for the clinical foundation of radioembolization, patient selection, and liver-directed therapy.
- Transarterial Radioembolization (TARE) / Selective Internal Radiation Therapy (SIRT) for the Treatment of Malignant Cancers of the Liver — UHC provider portal
- Radioembolization for Primary and Metastatic Tumors of the Liver — medical policy overview
- Advances and Emerging Techniques in Y-90 Radioembolization — broader review of evolving techniques
- Multimodal imaging techniques in Yttrium-90 radioembolization for hepatocellular carcinoma — strong review of imaging in Y-90 workflows
- Internal dosimetry for radioembolization therapy with Yttrium-90 microspheres — good for understanding dose logic and dosimetric thinking
2. AI in radiotherapy and theranostics
Best for understanding how AI is reshaping planning, prediction, and treatment workflows.
- The Role of Artificial Intelligence in Theranostics — practical and relevant overview
- AI-driven transformation of precision medicine: a comprehensive narrative review — broader AI context
- A deep-learning-based prediction model for the biodistribution of 90Y microspheres in liver radioembolization — especially relevant to your section on prediction
- Using Deep Learning to Predict Treatment Response in Patients with Hepatocellular Carcinoma Treated with Y90 Radiation Segmentectomy
- Automating RT Planning at Scale: High Quality Data For AI Training — useful for the infrastructure and dataset side of AI adoption
3. Dosimetry and computational planning
For the technical bridge between classical planning and next-generation AI-supported dosimetry.
- Optimizing dosimetry in Y-90 microsphere radioembolization
- Deep Learning-based Dosimetry Model for Patient-Specific 90Y
- Personalised Dosimetry in Nuclear Medicine: Bridging Physics, Biology and AI for Next Generation Radiopharmaceutical Therapy
- Demo: Generative AI helps Radiotherapy Planning with User Preference — interesting for interactive planning concepts
4. Combined modalities, theranostic pairs, and immunotherapy
Best for your section on synergy between Y-90, external radiation, and targeted molecular frameworks.
- Study Finds No Increase in Toxicity with Combined Radiation — relevant to sequential therapy safety
- 68Gallium- and 90Yttrium-/177Lutetium: “theranostic twins” for diagnosis and treatment of NETs
- 86/90Y-Based Theranostics Targeting Angiogenesis in a Murine Breast Cancer Model
- 68Ga/177Lu-Labeled Theranostic Pair for Targeting Fibroblast Activation Protein with Improved Tumor Uptake and Retention
- Radiotherapy + Immunotherapy for Liver Cancer — useful for the immunotherapy direction
5. Regulation, validation, and standards
Important if you want the paper to sound serious and policy-aware.
- A joint ESTRO and AAPM guideline for development, clinical validation and reporting of artificial intelligence models in radiation therapy
- ESTRO/AAPM Unit for Machine Learning Applications for Radiotherapy Outcome Modeling (UN99)
- A Joint ESTRO and AAPM Guideline for Development, Clinical Validation and Reporting of Artificial Intelligence Models in Radiation Therapy — alternate access version
6. Global infrastructure, access, and isotope production
For your LMIC and global access section.
- Rays of Hope — IAEA
- Rays of Hope Forum — IAEA
- Nuclear Science and Technology Delivers Hope to Cancer Patients Around the World — IAEA
- Production Review of Accelerator-Based Medical Isotopes
- Production of novel diagnostic radionuclides in small medical cyclotrons
- Production of radioisotopes for cancer imaging and treatment with compact linear accelerators
- The CNIC and IAEA highlight Canada’s first contributions — useful for the Canada angle