Anirudh Nimmagadda
Building at the interface of medicine & artificial intelligence.
Building
Damaros started with clinical trial execution: helping a community clinic run a trial as rigorously as a major cancer center, so where a patient lives stops deciding what care they can reach.
Pilots pointed past trials to a bigger bottleneck in medicine. That’s what I’m building towards now.
Games
Medicine, made playable.
Ganform
Draw the line, defend the organ.
A free, open-source contouring game on real CT. A daily puzzle, full-scan challenges where you fix auto-contours with planted errors, and Beam Puzzle, where you aim radiation beams at a tumor and keep the dose off the organs around it. Scored against public reference contours. Everything runs in the browser and nothing you draw is uploaded.
ganform.com For education, not a medical device.
The party game that speaks fluent medicine.
A free, phone-friendly multiplayer party game about medicine, for anyone. Everyone plays in the browser, no app or account. Humor as a healthy way to cope with it all: the hours, the jargon, the absurdity of the system.
iatrome.com Coming soon.
Seven nights, one doctor, the system never sleeps.
A medical horror game about night float. You watch five stations, and the queues do not wait for each other. Pagers won't stop, orders fall through, errors go in your file, and the footsteps get closer. Built in Godot.
In progress.
Research
Medicine, reinvestigated.
Selected workRadiation, cancer and AI8
- Target-Positive but Underdosed: Genomic-Adjusted Absorbed-Dose Requirements for Radiopharmaceutical TherapySubmitted, ASTRO 2027 RPT SymposiumTarget expression does not predict radiosensitivity, and fixed activity underdoses 96% of SSTR2-high paragangliomas.
- Measured, Not Prescribed: Dosimetry, Fixed-Activity Labels and Registry Reporting in 571 Radiopharmaceutical Therapy TrialsSubmitted, ASTRO 2027 RPT SymposiumDosimetry planning rose to 54% of trials, yet about 3% use it to set a patient's activity.
- Trial Access After Radiopharmaceutical Therapy: A Registry Analysis of 20,637 Treatment TrialsSubmitted, ASTRO 2027 RPT Symposium98% of phase 2 and 3 trials say nothing about prior RPT; where they do, one in five provisions excludes treated patients outright.
- PSA digital twins flag resistance to androgen receptor pathway inhibitors by day 60, including in men whose PSA has already halvedSubmitted, AUA 2027The twin beat day-60 PSA response (C-index 0.82 vs 0.76) and flagged one in six apparent responders who fared like nonresponders.
- Serious Pneumonitis Across 5,998 Thoracic Oncology Trial Arms: Registry-Wide Risk by Regimen and Radiotherapy Setting, and Projected Medicare BurdenSubmitted, ATS 2027Highest with chemoradiation plus a checkpoint inhibitor (6.4%) against 1.7% with SBRT plus one; about 2,900 Medicare cases a year.
- Who still implants the prostate? The shrinking prostate brachytherapy workforce and geographic access in Medicare, 2013-2024Submitted, AUA 2027Clinicians billing implants fell 58%; only 43% of Americans live within an hour's drive of a regular implanter.
- Synthetic Patient Data Generation with WGAN-GP and Mahalanobis Distance FilteringPublished, AIMH Austrian Journal, 2025Synthetic Alzheimer's patient records from a Wasserstein GAN, filtered by Mahalanobis distance so only realistic records are kept.
- Primary Care Datasets Utilized in AI and Machine LearningIn preparation, Stanford HEA₃RT LabSystematic review with the Stanford HEA₃RT Lab; 391 studies screened across title, abstract and full text.
All work, by specialty
Radiation Oncology and RadiobiologyASTRO, AUA, ATS, CNS10
- Target-Positive but Underdosed: Genomic-Adjusted Absorbed-Dose Requirements for Radiopharmaceutical TherapySubmitted, ASTRO 2027 RPT SymposiumTarget expression does not predict radiosensitivity, and fixed activity underdoses 96% of SSTR2-high paragangliomas.
- Measured, Not Prescribed: Dosimetry, Fixed-Activity Labels and Registry Reporting in 571 Radiopharmaceutical Therapy TrialsSubmitted, ASTRO 2027 RPT SymposiumDosimetry planning rose to 54% of trials, yet about 3% use it to set a patient's activity.
- Trial Access After Radiopharmaceutical Therapy: A Registry Analysis of 20,637 Treatment TrialsSubmitted, ASTRO 2027 RPT Symposium98% of phase 2 and 3 trials say nothing about prior RPT; where they do, one in five provisions excludes treated patients outright.
- Who still implants the prostate? The shrinking prostate brachytherapy workforce and geographic access in Medicare, 2013-2024Submitted, AUA 2027Clinicians billing implants fell 58%; only 43% of Americans live within an hour's drive of a regular implanter.
- Do prostate radiotherapy trials follow patients long enough to capture late urinary toxicity? A registry analysis of 613 trialsSubmitted, AUA 2027Half stop toxicity follow-up before five years, capturing about 45% of the urinary toxicity expected by ten.
- Endobronchial Brachytherapy in Decline: Medicare Use 2013-2024 and a Dose-Harmonized Synthesis of Fatal or Massive HemoptysisSubmitted, ATS 2027Medicare use fell from 396 services to 44, while randomized evidence shows no clear hemoptysis excess from the brachytherapy itself.
- Radiobiological Debulking: Gross Total Resection Targets Glioblastoma's Hypoxic Radioresistant NicheIn preparation, CNS 2027Radioresistance concentrates in the perinecrotic core surgery removes; residual enhancing volume tracked with survival (HR 1.96 per SD).
- Geographic Access to Actively Recruiting Radiotherapy Trials Relative to Radiotherapy Facility Access in the United StatesIn preparation, ASTRO 202732.7 million Americans live within an hour of radiotherapy but not of a recruiting radiotherapy trial; over a third in nonmetropolitan areas.
- Premarket Evidence and Postmarket Safety of Artificial Intelligence Devices Authorized for RadiotherapyIn preparation, ASTRO 2027Fewer than half of 89 FDA-cleared radiotherapy AI devices report a sized patient test set; one was tested prospectively.
- Genomic and Clinical Machine Learning Prediction of Brain Metastasis After Diagnosis of Non-Metastatic Non-Small Cell Lung CancerIn preparation, ASTRO 2027STK11, TP53 and CDKN2A alterations predicted brain metastasis; the top-risk quintile had an 18.5% two-year risk.
Thoracic OncologyATS2
- Serious Pneumonitis Across 5,998 Thoracic Oncology Trial Arms: Registry-Wide Risk by Regimen and Radiotherapy Setting, and Projected Medicare BurdenSubmitted, ATS 2027Highest with chemoradiation plus a checkpoint inhibitor (6.4%) against 1.7% with SBRT plus one; about 2,900 Medicare cases a year.
- Interstitial Lung Disease on United States Cancer Death Certificates, 1999-2024: A Cancer-Specific Rise Coinciding With the Immunotherapy EraSubmitted, ATS 2027Flat until 2015, then rising 5.9% a year: about 1,000 excess ILD-listed cancer deaths in 2024.
Urologic OncologyAUA2
- PSA digital twins flag resistance to androgen receptor pathway inhibitors by day 60, including in men whose PSA has already halvedSubmitted, AUA 2027The twin beat day-60 PSA response (C-index 0.82 vs 0.76) and flagged one in six apparent responders who fared like nonresponders.
- Do trials of novel systemic therapy for muscle-invasive bladder cancer let patients keep their bladder? An analysis of 234 registered trialsSubmitted, AUA 2027Only 39% admit bladder-preserved patients, and all three pivotal phase 3 trials required cystectomy.
Medical OncologyASCO2
- Pharmacovigilance of radiation-specific toxicities with systemic anticancer agents: A radiation-restricted disproportionality analysis of FAERSIn preparation, ASCO 2027Restricting to radiation-exposed reports cut flagged signals from 92% to 37%, and flags T-DXd and tucatinib for radiation necrosis.
- Targetable genotype and end-of-life oral, infusional and radiation therapy in a natural language processing-curated clinicogenomic cohortIn preparation, ASCO 202765% of systemic therapy in the last 14 days of life was oral only; a targetable genotype raised oral-targeted use near death (aOR 4.82).
AI and DataAIMH, Stanford2
- Synthetic Patient Data Generation with WGAN-GP and Mahalanobis Distance FilteringPublished, AIMH Austrian Journal, 2025Synthetic Alzheimer's patient records from a Wasserstein GAN, filtered by Mahalanobis distance so only realistic records are kept.
- Primary Care Datasets Utilized in AI and Machine LearningIn preparation, Stanford HEA₃RT LabSystematic review with the Stanford HEA₃RT Lab; 391 studies screened across title, abstract and full text.
DermatologyAAD2
- Xylazine-Associated Wound Complications and Dermatologic Artificial Intelligence: A Scoping Review of Current Evidence and Opportunities for Standardized Wound AssessmentSubmitted, AAD 2027Of 12 studies, none applied AI to xylazine wounds, and the wound criteria themselves are not yet standardized.
- Hidradenitis Suppurativa and Artificial Intelligence: A Scoping Review of Current Evidence and Opportunities for Earlier Diagnosis and Standardized Severity AssessmentSubmitted, AAD 2027Across 21 publications, no prospectively validated, image-based, lesion-level severity tool exists yet.
AnesthesiologyASRA2
- Rebound Pain After Peripheral Nerve Block: A Scoping Review of Current Evidence and Opportunities for AI-Enabled Risk StratificationSubmitted, ASRA 2027Rebound pain affects roughly 40 to 60% of block patients, and the only dedicated prediction model performs poorly (AUROC 0.632).
- Gastric Point-of-Care Ultrasound for Aspiration Risk Assessment Before Regional Anesthesia and Pain Procedures: A Scoping Review of Indications, Training, and Implementation BarriersIn preparationThe evidence supports selective, not routine, gastric POCUS when aspiration risk is high, such as in labor, diabetes or GLP-1 RA use.
OrthopaedicsORS2
- Bridge-Enhanced ACL Repair: A Scoping Review of Current Evidence and Opportunities for Standardization in Patient Selection, Failure Definitions, and Return-to-Sport ReportingAccepted, ORS 2027Encouraging in selected acute tears, but failure and return-to-sport definitions are too inconsistent to compare with reconstruction.
- Early WBAT vs NWB in Nonoperative Jones Fractures: A Scoping Review of Outcomes in Athletic and Non-Athletic PopulationsPresented, ORS 2026Early weight-bearing appears as safe as non-weight-bearing in non-athletes; data in athletes are sparse.
Background
UT San AntonioLong School of Medicine, MD/MSAI2024 to 2029
- The first dual MD and master’s in AI program. Co-president of the AI in Medicine Student Interest Group, 2025.
Stanford School of MedicineHEA₃RT Lab2025
- AI in population health and primary care: dataset bias, prediction and safety.
UT SouthwesternAbrams Lab2021 to 2024
- p53 and LINE-1 cancer biology in Drosophila and mouse models.
Baylor Scott & WhiteHospital Volunteer2021 to 2024
- Emergency department, 2021 to 2023. Operating room, 2023 to 2024.
UT DallasBS in Biology2021 to 2023
- Founded the UT Dallas chapter of AI Student Collective, 2023 to 2024.