The work, led by Jheremy S. Reyes, Constantinos G. Hadjipanayis, L. Dade Lunsford, and Ajay Niranjan at the University of Pittsburgh Medical Center’s Center for Image-Guided Neurosurgery, addresses a population that has long been underserved by predictive modeling. Brain metastases from gastrointestinal primaries, including colorectal, pancreatic, gastric, and biliary cancers, are biologically heterogeneous and historically associated with poorer local control than metastases from lung or breast cancers. Because these tumors are relatively less common in the radiosurgery literature, dose selection for them has been extrapolated from broader metastasis cohorts rather than tailored to their specific behavior. The Pittsburgh team set out to build a model trained specifically on this patient group.

At the heart of the framework is a machine learning architecture known as a mixture of experts, a concept that dates back to foundational work by Jacobs, Jordan, Nowlan, and Hinton in 1991 and was later formalized hierarchically by Jordan and Jacobs. Rather than forcing a single neural network to learn every pattern in the data, a mixture-of-experts model divides the problem among several specialized subnetworks, each of which becomes adept at a particular region of the clinical feature space. A gating network then learns which expert to trust for a given patient, weighting their contributions accordingly. In THINKERS-GI, three such experts operate in parallel, allowing the model to capture distinct subpopulations among GI brain metastasis patients whose tumors and outcomes behave in fundamentally different ways.

The technical machinery beneath the experts is equally deliberate. The framework employs discrete-time competing-risk survival modeling, a statistical approach that estimates the probability of an event, such as local tumor failure, over successive time intervals while explicitly accounting for competing events that can preclude that failure, most notably death from systemic disease. This matters enormously in brain metastases, where many patients have limited life expectancies driven by their systemic cancer, and a model that ignores competing risks will systematically overestimate the benefit of aggressive local therapy. By modeling both local failure and adverse radiation effects in this competing-risk framework, THINKERS-GI can express its outputs as time-dependent probabilities rather than crude binary classifications.

One of the most clinically significant design choices is that the margin dose, the radiation dose prescribed at the edge of the tumor, was built into the model as a queryable input. This transforms the network from a passive outcome predictor into an active decision-support tool. A clinician can ask the model a counterfactual question: what would the predicted probability of local control and of adverse radiation effects be if this lesion received 18 Gy instead of 16 Gy? By sweeping across a range of candidate doses, the framework generates a dose-response profile unique to each lesion, effectively allowing the care team to explore the trade-off between tumor control and toxicity before the first beam is ever delivered.

The study drew on a retrospective single-center cohort of 300 patients treated with Gamma Knife radiosurgery, encompassing 1,233 individual GI brain metastases. Pretreatment clinical and lesion-level variables were used to train the network, and the authors took validation seriously, employing grouped five-fold cross-validation alongside a closed, grouped holdout set. Grouping is a crucial methodological safeguard in this setting: because a single patient can contribute many lesions, randomly splitting lesions across training and test sets would leak patient-specific information and inflate performance estimates. By keeping all lesions from one patient together, the researchers ensured that their reported metrics reflect genuine generalization to unseen patients.

The results were strong by the standards of clinical prediction modeling. In grouped cross-validation, the time-dependent area under the curve ranged from 0.808 to 0.894, with a concordance index of 0.889 plus or minus 0.034, indicating good discrimination between patients who experienced local failure and those who did not. The calibrated twelve-month Brier score, which penalizes both poor discrimination and poor calibration, was a remarkably low 0.019 plus or minus 0.013, and the mean absolute error for predicting the time to local failure was just 1.8 plus or minus 0.7 months. On the more demanding grouped holdout set, performance remained respectable, with a twelve-month AUC of 0.806, a C-index of 0.781, a calibrated Brier score of 0.101, and a mean absolute error for time-to-failure of 1.27 months.

The dose-sweeping experiments offer a glimpse of what routine clinical use might look like. When the framework evaluated candidate doses across the cohort, it selected a dose within two Gy of the dose actually delivered by physicians in 91.8 percent of lesions, a striking signal that the model’s reasoning converges with expert clinical judgment in the vast majority of routine cases. More intriguingly, in the cases where the model suggested deviations, the simulated optimization increased the predicted twelve-month local control rate from 82.3 percent to 84.4 percent while simultaneously reducing the predicted risk of adverse radiation effects from 7.9 percent to 7.6 percent. Improving both endpoints at once is the holy grail of radiation dose optimization, since these outcomes typically pull in opposite directions.

The THINKERS-GI study is not an isolated effort but part of a rapidly expanding research program from the Pittsburgh group, which has applied the same hybrid intelligence philosophy to meningiomas, glioblastomas, vestibular schwannomas, lung cancer brain metastases, melanoma brain metastases, pituitary adenomas, craniopharyngiomas, and post-resection cavities. This disease-specific strategy reflects a growing recognition in oncology AI that a single general-purpose model often performs worse than a family of specialized ones, particularly for rare tumor types with distinctive biology. Related work elsewhere, such as the MoE-Surv framework for rare cancers, has reached similar conclusions, suggesting that mixture-of-experts architectures may become a standard tool for survival prediction in oncology.

Important caveats remain before such tools reach the treatment planning console. The study was retrospective and single-center, and the authors describe their work as internally validated, meaning the model has not yet been tested on patients from other institutions with different treatment protocols and patient populations. External validation, and eventually prospective evaluation, will be essential to establish that the framework’s dose recommendations translate into genuinely better outcomes rather than merely better predictions. The authors also note that no external datasets were generated or analyzed in the current study, underscoring that this is a foundation rather than a finished clinical product. Still, the combination of rigorous competing-risk survival modeling, patient-grouped validation, and queryable dose inputs marks a meaningful step toward a future in which every radiosurgery prescription is informed by an individualized estimate of what the treatment is likely to achieve, and what it is likely to cost the surrounding brain.

Subject of Research: A mixture-of-experts AI framework for predicting outcomes and optimizing radiation doses in gastrointestinal brain metastases treated with Gamma Knife radiosurgery

Article Title: Therapeutic hybrid intelligence with neural and knowledge-based expert reasoning for SRS (THINKERS): a mixture-of-experts framework for gastrointestinal brain metastases

Article References: Reyes, J. S., Hadjipanayis, C. G., Lunsford, L. D., & Niranjan, A. (2026). Therapeutic hybrid intelligence with neural and knowledge-based expert reasoning for SRS (THINKERS): a mixture-of-experts framework for gastrointestinal brain metastases. Journal of Neuro-Oncology, 179(2), Article 68. https://doi.org/10.1007/s11060-026-05781-3

Image Credits: AI Generated

DOI: 10.1007/s11060-026-05781-3

Keywords: brain metastases, gastrointestinal cancer, Gamma Knife radiosurgery, artificial intelligence, mixture of experts, dose optimization, survival modeling, local control, radiation toxicity, predictive modeling, neuro-oncology, competing risks

Pilocytic astrocytoma is a WHO grade 1 glioma, a tumor class that behaves far more indolently than the aggressive glioblastomas that dominate public awareness of brain cancer. These tumors are driven largely by alterations in the MAPK signaling pathway and, in children, are often cured by surgery alone. In adults, however, the disease is less well characterized. Adult tumors tend to arise in different locations, may be less amenable to complete resection, and can recur years or even decades after an initial operation. When they sit in the brainstem, the optic pathway, the hypothalamus, or deep midline ventricular structures, the risk of repeated surgery becomes unacceptable, leaving radiotherapy as the principal local option.

The Heidelberg team, led by Fabian J. K. Allmendinger and colleagues, assembled a retrospective cohort of fourteen adults treated with intensity-modulated proton radiotherapy between 2011 and 2025. The patients were young, with a median age of 23.5 years at treatment, and their tumors occupied some of the most anatomically treacherous territory in the brain: the cerebellum, the frontal and temporal lobes, the third and fourth ventricles, the suprasellar compartment, the mesencephalon, and the pineal region. Ten of the fourteen had undergone prior surgery, and in nine of those cases the operation had achieved only subtotal removal. For ten patients, proton therapy was salvage treatment for recurrent disease; for four, it was the primary local therapy when surgery was judged too risky.

The technical premise of proton therapy is elegantly simple. Unlike conventional photon beams, which deposit radiation along their entire path through tissue and exit on the far side of the target, protons carry a finite range. They release most of their energy in a sharply defined burst, the Bragg peak, at a depth determined by their initial energy, and deliver essentially nothing beyond it. By scanning pencil-thin proton beams across the tumor volume, clinicians can conform the dose to irregular targets while dramatically reducing the radiation bath received by uninvolved brain, the temporal lobes, the hippocampi, the optic apparatus, the hypothalamic-pituitary axis, and cerebral blood vessels. For patients expected to live for decades, that reduction in integral dose is not a cosmetic advantage; it is the difference between preserving and progressively eroding neurocognitive, endocrine, and visual function over a lifetime.

The prescribed dose in the cohort was a median of 54.0 Gy(RBE), delivered in roughly 29 fractions, with planning that co-registered treatment CT scans with contrast-enhanced T1-weighted and T2/FLAIR MRI and applied QUANTEC-based constraints for organs at risk. What followed was, by the standards of brain tumor therapy, remarkable. At last follow-up, bidirectional T2-weighted tumor measurements showed complete regression in three patients, partial response of at least fifty percent in seven, and a minor response in two. Only two patients experienced progressive disease. Perhaps most intriguingly, shrinkage frequently continued beyond the twelve-month landmark: ten of the twelve non-progressive patients kept improving after the standardized one-year assessment, a pattern consistent with the slow growth kinetics of grade 1 tumors and a warning against prematurely declaring treatment failure.

Survival outcomes matched the radiographic enthusiasm. Estimated overall survival was one hundred percent at both three and five years, with no deaths recorded within the first five years after therapy. Progression-free survival stood at 83.3 percent at both timepoints, and most progression events occurred early, meaning that patients who remained disease-free in the initial years tended to stay that way. Median clinical follow-up extended to nearly ten years, with vital status confirmed through national registry data, lending unusual depth to the survival analysis for a cohort of this size.

Toxicity was predominantly low grade. During treatment, the most common acute effects were focal alopecia in 64 percent of patients, fatigue and headache in 43 percent each, and radiodermatitis in 29 percent, and nearly all of these resolved in follow-up. Alopecia and skin reactions vanished entirely beyond the first months. Fatigue was the most persistent complaint, affecting three patients across all follow-up intervals, while mild cognitive impairment appeared in two. One patient with pre-existing postoperative visual deficits developed late grade 1 visual impairment, though optic nerve dose constraints had been met and imaging showed no optic neuropathy. Critically, no grade 4 or grade 5 events occurred at any point.

The study also shines a light on a subtle diagnostic trap known as radiation-induced contrast enhancement, or RICE. Within months to years after proton therapy, new areas of contrast uptake can appear on MRI within or near the irradiated volume, mimicking tumor progression on a scan. In this cohort, three patients developed RICE, yielding a three- and five-year RICE-free probability of 76.2 percent, with all events occurring in the early post-treatment period. One patient was initially misclassified as having progressed and was started on temozolomide chemotherapy, which had to be abandoned after a single cycle due to blood toxicity; retrospective review showed the enhancing lesion was radiation injury, not tumor, and bevacizumab sufficed. The authors emphasize that distinguishing RICE from true progression, using imaging morphology, temporal evolution, spatial relation to the dose distribution, and clinical course, is essential to avoid unnecessary chemotherapy in patients who otherwise face decades of life.

The findings arrive amid an evolving treatment landscape. Current EANO-EURACAN-SNO guidelines recommend maximal safe resection when feasible, with molecular testing for BRAF and MAPK alterations, and acknowledge reoperation, radiotherapy, and targeted agents such as BRAF and MEK inhibitors as options for recurrent or unresectable tumors. Recent multicenter data show chemotherapy with temozolomide achieves a median progression-free survival of only about twenty months in adult pilocytic astrocytoma, a figure that looks modest against the durable control seen here. The Heidelberg authors are careful, however, not to overreach: the study was retrospective, involved only fourteen patients, lacked a photon comparator, and carried incomplete molecular annotation, so it cannot prove that protons outperform photons, only that they deliver excellent control with acceptable morbidity in a selected, anatomically complex population.

Even with those caveats, the message for patients and clinicians is consequential. For adults facing a progressive pilocytic astrocytoma that surgery cannot safely reach, proton radiotherapy now has dedicated, adult-specific evidence behind it: near-universal radiographic response, complete five-year survival, and side effects that were mostly transient and mild. Because prospective trials are unlikely in a disease this rare, single-center cohorts like this one carry disproportionate weight. The study also reinforces a broader principle of modern radiation oncology, that in long-surviving patients with benign-behaving tumors, the true measure of a treatment is not only whether it controls the disease, but what quality of brain, and of life, it leaves behind ten, twenty, and thirty years later.

Subject of Research: Proton radiotherapy for progressive or surgically unfavorable adult pilocytic astrocytoma

Article Title: Efficacy and safety of proton radiotherapy for progressive or surgically unfavorable adult pilocytic astrocytoma: a single-center experience

Article References: Allmendinger, F. J. K., Cherniienko, Y., Lischalk, J., Walter, J., Mose, L., Wickert, R., Deng, M., Regnery, S., Wessel, L., Kozyra, K., Tessonnier, T., Krieg, S., Debus, J., König, L., & Adena-Eichkorn, T. (2026). Efficacy and safety of proton radiotherapy for progressive or surgically unfavorable adult pilocytic astrocytoma: a single-center experience. Journal of Neuro-Oncology, 179(3), Article 89. https://doi.org/10.1007/s11060-026-05798-8

Image Credits: AI Generated

DOI: 10.1007/s11060-026-05798-8

Keywords: proton radiotherapy, pilocytic astrocytoma, adult glioma, radiation oncology, RICE, tumor control, low-grade glioma, Bragg peak, neuro-oncology, radiation toxicity, MAPK pathway, brain tumors