Model Medicine is the science of understanding, diagnosing, treating, and preventing disorders in AI models, grounded in the principle that AI models -- like biological organisms -- have internal structures, dynamic processes, heritable traits, observable symptoms, classifiable conditions, and treatable states. This paper introduces Model Medicine as a research program, bridging the gap between current AI interpretability research (anatomical observation) and the systematic clinical practice that complex AI systems increasingly require. We present five contributions: (1) a discipline taxonomy organizing 15 subdisciplines across four divisions -- Basic Model Sciences, Clinical Model Sciences, Model Public Health, and Model Architectural Medicine; (2) the Four Shell Model (v3.3), a behavioral genetics framework empirically grounded in 720 agents and 24,923 decisions from the Agora-12 program, explaining how model behavior emerges from Core--Shell interaction; (3) Neural MRI (Model Resonance Imaging), a working open-source diagnostic tool mapping five medical neuroimaging modalities to AI interpretability techniques, validated through four clinical cases demonstrating imaging, comparison, localization, and predictive capability; (4) a five-layer diagnostic framework for comprehensive model assessment; and (5) clinical model sciences including the Model Temperament Index for behavioral profiling, Model Semiology for symptom description, and M-CARE for standardized case reporting. We additionally propose the Layered Core Hypothesis -- a biologically-inspired three-layer parameter architecture -- and a therapeutic framework connecting diagnosis to treatment.
Abstract Background In-silico and in-vitro studies have revealed an appropriate posterior tibial slope (PTS) is critical for normal anterior cruciate ligament (ACL) and posterior cruciate ligament (PCL) tension and knee biomechanical behavior of unicompartmental knee arthroplasty (UKA). However, the effects of PTS on in-vivo elongation of ACL and PCL in UKA remains unknown. The study aimed to quantify in-vivo ACL and PCL elongations during lunge and analyze their relations with PTS. Methods Thirteen fixed-bearing (FB) and 11 mobile-bearing (MB) UKA patients were recruited. The postoperative medial PTS was defined as the angle between the tibial transverse plane (perpendicular to mechanical axis) and cut plane. Accurate knee spatial postures of UKA and contralateral native knees during single-leg lunge were measured by the dual fluoroscopic imaging system. The ACL (AM, PL bundles) and PCL (AL, PM bundles) footprints were determined based on anatomical features on femoral and tibial 3D surface model reconstructed from CT. A validated 3D wrapping method was used to measure ligament bundle length. The paired Wilcoxon signed-rank test was used to analyze the ligament elongation difference between bilateral knees. The Spearman correlation between PTS and average ligament elongation difference (ACL during 0–30° early-flexion, PCL during 60–100° deep-flexion) was calculated. Results The elongation of FB UKA PCL double-bundle was larger than contralateral sides in most flexion range of lunge (Max-Difference: AL 7.6 ± 8.7%, PM 8.2 ± 5.1%, p < 0.05). In contrast, ACL double-bundle elongations of MB UKA in mid-flexion were larger than contralateral sides (Max-Difference: AM 8.0 ± 8.1%, PL 7.6 ± 9.8%, p < 0.05). The increased PTS was significantly relevant to the increased ACL double-bundle elongation difference of bilateral knees for both FB and MB UKA patients (R > 0.6, p < 0.05). Conclusion There was abnormal in-vivo elongation of PCL in FB UKA and ACL in MB UKA during lunge and cause over-constraints to the contralateral knee. There was a positive correlation between PTS and ACL elongation difference for both FB and MB UKA, indicating excessive PTS should be avoided to preserve native ACL function in further UKA implantation. Levels of Evidence III.
The aim of this study was to investigate the role of point of care ultrasound (POCUS) as an alternative imaging modality to confirm the location of gastric and post-pyloric feeding tubes in patients admitted to the pediatric intensive care unit (PICU). This was a prospective descriptive study performed at a tertiary care children’s hospital. Patients from birth to 17 years of age in whom the medical team placed a temporary enteral feeding tube were eligible for enrollment. The study physician, who was blinded to the radiographic findings, performed a POCUS study of the abdomen. An abdominal radiograph was obtained to confirm the placement in all patients. A total of 13 patients were enrolled, and 14 abdominal POCUS exams were completed. POCUS accurately identified the location of the enteral feeding tube in 10 of the 14 cases. POCUS had a sensitivity and specificity of 85.7% and 57.1%, respectively, in identifying gastric tubes. It had a sensitivity and specificity of 66.7% and 87.5%, respectively, in identifying post-pyloric tubes. No adverse events were reported. This study showed that POCUS had moderate sensitivity and specificity and was, overall, safe. Further studies can assess the level of training needed for improvement in accuracy, and larger studies can help support the findings of this data that POCUS is a safe and accurate alternative to radiographs for enteral feeding tube placement confirmation.
Jean S. Edward, Amanda Thaxton Wiggins, Louis G. Baser
et al.
Few evidence-based trainings exist on how to equip healthcare providers, particularly nurses, with the skills to engage in cost of care conversations with patients/caregivers to mitigate the impact of cancer-related financial toxicity. This study evaluated a pilot training developed in collaboration with Triage Cancer<sup>®</sup> to prepare oncology nurses to identify and assist patients/caregivers facing financial and/or legal barriers to care. Ten pediatric oncology nurses completed the training and pre/post-surveys on behaviors related to financial and legal need screening, frequency and comfort level of answering questions, knowledge, and behavior changes, along with training evaluation questions. At baseline, six nurses reported never screening for financial needs and nine for legal needs. Following the training, seven nurses stated they were likely to screen for financial/legal needs. At six months post-training, nurses had referred 85 patients/caregivers to financial/legal navigation services. Comfort levels in answering financial/legal questions increased by 6.5 points and knowledge scores increased by 1.7 points post-training. Most nurses recommended this training to other healthcare providers who work with patients with cancer and their caregivers. This study highlights the importance of providing oncology nurses with resources to engage in cost of care conversations and oncology financial legal navigation programs to mitigate the impact of cancer-related financial toxicity.
Neoplasms. Tumors. Oncology. Including cancer and carcinogens
Laura Izquierdo Sanchez, Julen Matin Robles, Jone Narbaiza
et al.
Introduction and Objectives: Cholangiocarcinoma (CCA) incidence and mortality are rising globally. Chronic liver diseases (CLD) are recognized risk factors. This study aimed to compare the clinical presentation and outcomes of CCA in patients with and without CLD, using data from the International CCA Registry. Patients and Methods: The international CCA Registry is a multicenter observational study enrolling cases from 54 centers across Latin America, Europe, and Asia (2010–2024). Results: Among 3,693 patients enrolled, 916 had CLD and 2,777 did not. Common CLD conditions were fatty liver disease, cirrhosis, viral hepatitis, and primary sclerosing cholangitis. Compared to non-CLD patients, those with CLD were more often male (69% vs. 53%), younger at diagnosis (63 vs. 66 years), and had higher rates of metabolic risk factors, alcohol use, and smoking. Intrahepatic CCA was more frequent in CLD patients (64% vs. 43%), whereas distal CCA was more common in non-CLD cases (20% vs. 9%). CLD patients had better performance status (ECOG 0: 53% vs. 35%), lower CA19-9 levels (59.0 vs. 134.5 U/mL), and more localized disease (56% vs. 48%). Curative-intent surgery was more frequent in the CLD group (59% vs. 48%), translating into longer median overall survival (12.3 vs. 11.0 months) and higher 5-year survival (OR = 1.67; p < 0.001). The benefit was especially evident in intrahepatic CCA. Treatment responses were comparable between groups. Conclusions: CCA is diagnosed at earlier stages in individuals with CLD, likely due to certain clinical surveillance, leading to better prognosis. Prospective validation and standardized surveillance protocols are warrant.
Periklis Petridis, Georgios Margaritis, Vasiliki Stoumpou
et al.
With the increasing interest in deploying Artificial Intelligence in medicine, we previously introduced HAIM (Holistic AI in Medicine), a framework that fuses multimodal data to solve downstream clinical tasks. However, HAIM uses data in a task-agnostic manner and lacks explainability. To address these limitations, we introduce xHAIM (Explainable HAIM), a novel framework leveraging Generative AI to enhance both prediction and explainability through four structured steps: (1) automatically identifying task-relevant patient data across modalities, (2) generating comprehensive patient summaries, (3) using these summaries for improved predictive modeling, and (4) providing clinical explanations by linking predictions to patient-specific medical knowledge. Evaluated on the HAIM-MIMIC-MM dataset, xHAIM improves average AUC from 79.9% to 90.3% across chest pathology and operative tasks. Importantly, xHAIM transforms AI from a black-box predictor into an explainable decision support system, enabling clinicians to interactively trace predictions back to relevant patient data, bridging AI advancements with clinical utility.
Robert A. Rigby, Mikis D. Stasinopoulos, Achim Zeileis
et al.
We read with interest the above article by Zavorsky (2025, Respiratory Medicine, doi:10.1016/j.rmed.2024.107836) concerning reference equations for pulmonary function testing. The author compares a Generalized Additive Model for Location, Scale, and Shape (GAMLSS), which is the standard adopted by the Global Lung Function Initiative (GLI), with a segmented linear regression (SLR) model, for pulmonary function variables. The author presents an interesting comparison; however there are some fundamental issues with the approach. We welcome this opportunity for discussion of the issues that it raises. The author's contention is that (1) SLR provides "prediction accuracies on par with GAMLSS"; and (2) the GAMLSS model equations are "complicated and require supplementary spline tables", whereas the SLR is "more straightforward, parsimonious, and accessible to a broader audience". We respectfully disagree with both of these points.
What does Artificial Intelligence (AI) have to contribute to health care? And what should we be looking out for if we are worried about its risks? In this paper we offer a survey, and initial evaluation, of hopes and fears about the applications of artificial intelligence in medicine. AI clearly has enormous potential as a research tool, in genomics and public health especially, as well as a diagnostic aid. It's also highly likely to impact on the organisational and business practices of healthcare systems in ways that are perhaps under-appreciated. Enthusiasts for AI have held out the prospect that it will free physicians up to spend more time attending to what really matters to them and their patients. We will argue that this claim depends upon implausible assumptions about the institutional and economic imperatives operating in contemporary healthcare settings. We will also highlight important concerns about privacy, surveillance, and bias in big data, as well as the risks of over trust in machines, the challenges of transparency, the deskilling of healthcare practitioners, the way AI reframes healthcare, and the implications of AI for the distribution of power in healthcare institutions. We will suggest that two questions, in particular, are deserving of further attention from philosophers and bioethicists. What does care look like when one is dealing with data as much as people? And, what weight should we give to the advice of machines in our own deliberations about medical decisions?
Objective To develop an LLM based realtime compound diagnostic medical AI interface and performed a clinical trial comparing this interface and physicians for common internal medicine cases based on the United States Medical License Exam (USMLE) Step 2 Clinical Skill (CS) style exams. Methods A nonrandomized clinical trial was conducted on August 20, 2024. We recruited one general physician, two internal medicine residents (2nd and 3rd year), and five simulated patients. The clinical vignettes were adapted from the USMLE Step 2 CS style exams. We developed 10 representative internal medicine cases based on actual patients and included information available on initial diagnostic evaluation. Primary outcome was the accuracy of the first differential diagnosis. Repeatability was evaluated based on the proportion of agreement. Results The accuracy of the physicians' first differential diagnosis ranged from 50% to 70%, whereas the realtime compound diagnostic medical AI interface achieved an accuracy of 80%. The proportion of agreement for the first differential diagnosis was 0.7. The accuracy of the first and second differential diagnoses ranged from 70% to 90% for physicians, whereas the AI interface achieved an accuracy rate of 100%. The average time for the AI interface (557 sec) was 44.6% shorter than that of the physicians (1006 sec). The AI interface ($0.08) also reduced costs by 98.1% compared to the physicians' average ($4.2). Patient satisfaction scores ranged from 4.2 to 4.3 for care by physicians and were 3.9 for the AI interface Conclusion An LLM based realtime compound diagnostic medical AI interface demonstrated diagnostic accuracy and patient satisfaction comparable to those of a physician, while requiring less time and lower costs. These findings suggest that AI interfaces may have the potential to assist primary care consultations for common internal medicine cases.
Arterial aneurysms remain a significant public health problem because they often result in death when ruptured; therefore, they require immediate medical treatment. Endovascular aneurysm repair (EVAR) has recently become the primary treatment option, owing to the fewer side effects compared to those with open surgery. However, stents used for conventional EVAR often cause side-branch occlusion, which alters the perfusion of vital organs. Recently, multilayer flow modulator (MFM) stents have been used as a new treatment for arterial aneurysms. These stents appear to be feasible owing to their unique design consisting of an uncoated three-dimensionally braided multilayered structure. MFM stents generally remodulate laminar flow and reduce the flow velocity in the aneurysmal sac, leading to thrombosis, which causes the aneurysm to shrink over time. Thus, they reduce the risk of mortality. Moreover, they reduce morbidity by preserving the side-branch blood flow. They can be easily applied to complex aneurysms and are ready to use without customization, which shortens the waiting time for interventions. This study aimed to evaluate the role of MFM stents in the treatment of arterial aneurysms based on available data.
Diseases of the circulatory (Cardiovascular) system
Matthias Christenson, Cove Geary, Brian Locke
et al.
The success of precision medicine requires computational models that can effectively process and interpret diverse physiological signals across heterogeneous patient populations. While foundation models have demonstrated remarkable transfer capabilities across various domains, their effectiveness in handling individual-specific physiological signals - crucial for precision medicine - remains largely unexplored. This work introduces a systematic pipeline for rapidly and efficiently evaluating foundation models' transfer capabilities in medical contexts. Our pipeline employs a three-stage approach. First, it leverages physiological simulation software to generate diverse, clinically relevant scenarios, particularly focusing on data-scarce medical conditions. This simulation-based approach enables both targeted capability assessment and subsequent model fine-tuning. Second, the pipeline projects these simulated signals through the foundation model to obtain embeddings, which are then evaluated using linear methods. This evaluation quantifies the model's ability to capture three critical aspects: physiological feature independence, temporal dynamics preservation, and medical scenario differentiation. Finally, the pipeline validates these representations through specific downstream medical tasks. Initial testing of our pipeline on the Moirai time series foundation model revealed significant limitations in physiological signal processing, including feature entanglement, temporal dynamics distortion, and reduced scenario discrimination. These findings suggest that current foundation models may require substantial architectural modifications or targeted fine-tuning before deployment in clinical settings.
In this paper, we introduce InMD-X, a collection of multiple large language models specifically designed to cater to the unique characteristics and demands of Internal Medicine Doctors (IMD). InMD-X represents a groundbreaking development in natural language processing, offering a suite of language models fine-tuned for various aspects of the internal medicine field. These models encompass a wide range of medical sub-specialties, enabling IMDs to perform more efficient and accurate research, diagnosis, and documentation. InMD-X's versatility and adaptability make it a valuable tool for improving the healthcare industry, enhancing communication between healthcare professionals, and advancing medical research. Each model within InMD-X is meticulously tailored to address specific challenges faced by IMDs, ensuring the highest level of precision and comprehensiveness in clinical text analysis and decision support. This paper provides an overview of the design, development, and evaluation of InMD-X, showcasing its potential to revolutionize the way internal medicine practitioners interact with medical data and information. We present results from extensive testing, demonstrating the effectiveness and practical utility of InMD-X in real-world medical scenarios.