Physics-informed deep learning sharpens nano diagnostics for elusive pancreatic cancer.

Real-World Surveillance of FDA-Cleared Artificial Intelligence Models: Rationale and Logistics.

Predictive modeling in reproductive medicine: Where will the future of artificial intelligence research take us?

Comment on “Usefulness of artificial intelligence for predicting recurrence following surgery for pancreatic cancer: Retrospective cohort study”

Rationale and design of the INVICTUS Registry: (Multicenter Registry of Invasive and Non-Invasive imaging modalities to compare Coronary Computed Tomography Angiography, Intravascular Ultrasound and Optical Coherence Tomography for the determination of Severity, Volume and Type of coronary atherosclerosiS)

Artificial intelligence (AI) has made significant advancements in the medical domain in recent years. AI, an expansive field comprising Machine Learning (ML) and, within it, Deep Learning (DL), seeks to emulate the intricate operations of the human brain. It examines vast amounts of data and plays a crucial role in decision-making, overcoming limitations related to human evaluation. DL utilizes complex algorithms to analyze data. ML and DL are subsets of AI that utilizehard statistical techniques that help machines consistently improve at tasks with experience. Pancreatic cancer is more common in developed countries and is one of the leading causes of cancer-related mortalityworldwide. Managing pancreatic cancer remains a challenge despite significant advancements in diagnosis and treatment. AI has secured an almost ubiquitous presence in the field of oncological workup and management, especially in gastroenterology malignancies. AI is particularly useful for various investigations of pancreatic carcinoma because it has specific radiological features that enable diagnostic procedures without the requirement of a histological study. However, interpreting and evaluating resulting images is not always simple since images vary as the disease progresses. Secondly, a number of factors may impact prognosis and response to the treatment process. Currently, AI models have been created for diagnosing, grading, staging, and predicting prognosis and treatment response. This review presents the most up-to-date knowledge on the use of AI in the diagnosis and treatment of pancreatic carcinoma.

An artificial intelligence (AI) model using chest radiography (CXR) may provide good performance in making prognoses for COVID-19. We aimed to develop and validate a prediction model using CXR based on an AI model and clinical variables to predict clinical outcomes in patients with COVID-19. This retrospective longitudinal study included patients hospitalized for COVID-19 at multiple COVID-19 medical centers between February 2020 and October 2020. Patients at Boramae Medical Center were randomly classified into training, validation, and internal testing sets (at a ratio of 8:1:1, respectively). An AI model using initial CXR images as input, a logistic regression model using clinical information, and a combined model using the output of the AI model (as CXR score) and clinical information were developed and trained to predict hospital length of stay (LOS) ≤2 weeks, need for oxygen supplementation, and acute respiratory distress syndrome (ARDS). The models were externally validated in the Korean Imaging Cohort of COVID-19 data set for discrimination and calibration. The AI model using CXR and the logistic regression model using clinical variables were suboptimal to predict hospital LOS ≤2 weeks or the need for oxygen supplementation but performed acceptably in the prediction of ARDS (AI model area under the curve [AUC] 0.782, 95% CI 0.720-0.845; logistic regression model AUC 0.878, 95% CI 0.838-0.919). The combined model performed better in predicting the need for oxygen supplementation (AUC 0.704, 95% CI 0.646-0.762) and ARDS (AUC 0.890, 95% CI 0.853-0.928) compared to the CXR score alone. Both the AI and combined models showed good calibration for predicting ARDS (P=.079 and P=.859). The combined prediction model, comprising the CXR score and clinical information, was externally validated as having acceptable performance in predicting severe illness and excellent performance in predicting ARDS in patients with COVID-19.

Addressing Characteristics, Response and Relapse: Eltrombopag As Third Line in Primary and Secondary Immune Thrombocytopenia

This study evaluates the diagnostic performance of several AI models, including Deepseek, in diagnosing corneal diseases, glaucoma, and neuro□ophthalmologic disorders. We retrospectively selected 53 case reports from the Department of Ophthalmology and Visual Sciences at the University of Iowa, comprising 20 corneal disease cases, 11 glaucoma cases, and 22 neuro□ophthalmology cases. The case descriptions were input into DeepSeek, ChatGPT□4.0, ChatGPT□01, and Qwens 2.5 Max. These responses were compared with diagnoses rendered by human experts (corneal specialists, glaucoma attendings, and neuro□ophthalmologists). Diagnostic accuracy and interobserver agreement, defined as the percentage difference between each AI model's performance and the average human expert performance, were determined. DeepSeek achieved an overall diagnostic accuracy of 79.2%, with specialty-specific accuracies of 90.0% in corneal diseases, 54.5% in glaucoma, and 81.8% in neuro□ophthalmology. ChatGPT□01 outperformed the other models with an overall accuracy of 84.9% (85.0% in corneal diseases, 63.6% in glaucoma, and 95.5% in neuro□ophthalmology), while Qwens exhibited a lower overall accuracy of 64.2% (55.0% in corneal diseases, 54.5% in glaucoma, and 77.3% in neuro□ophthalmology). Interobserver agreement analysis revealed that in corneal diseases, DeepSeek differed by -3.3% (90.0% vs 93.3%), ChatGPT□01 by -8.3%, and Qwens by -38.3%. In glaucoma, DeepSeek outperformed the human expert average by +3.0% (54.5% vs 51.5%), while ChatGPT□4.0 and ChatGPT□01 exceeded it by +12.1%, and Qwens was +3.0% above the human average. In neuro□ophthalmology, DeepSeek and ChatGPT□4.0 were 9.1% lower than the human average, ChatGPT□01 exceeded it by +4.6%, and Qwens was 13.6% lower. ChatGPT□01 demonstrated the highest overall diagnostic accuracy, especially in neuro□ophthalmology, while DeepSeek and ChatGPT□4.0 showed comparable performance. Qwens underperformed relative to the other models, especially in corneal diseases. Although these AI models exhibit promising diagnostic capabilities, they currently lag behind human experts in certain areas, underscoring the need for a collaborative integration of clinical judgment. This study evaluated how well several artificial intelligence (AI) models diagnose eye diseases compared to human experts. We tested four AI systems across three types of eye conditions: diseases of the cornea, glaucoma, and neuro-ophthalmologic disorders. Overall, one AI model, ChatGPT-01, performed the best, correctly diagnosing about 85% of cases, and it excelled in neuro-ophthalmology by correctly diagnosing 95.5% of cases. Two other models, DeepSeek and ChatGPT-4.0, each achieved an overall accuracy of around 79%, while the Qwens model performed lower, with an overall accuracy of about 64%. When compared with human experts, who achieved very high accuracy in corneal diseases (93.3%) and neuro-ophthalmology (90.9%) but lower in glaucoma (51.5%), the AI models showed mixed results. In glaucoma, for instance, some AI models even outperformed human experts slightly, while in corneal diseases, all AI models were less accurate than the experts. These findings indicate that while AI shows promise as a supportive tool in diagnosing eye conditions, it still needs further improvement. Combining AI with human clinical judgment appears to be the best approach for accurate eye disease diagnosis. Why carry out this study? With the rising burden of eye diseases and the inherent diagnostic challenges for complex conditions like glaucoma and neuro-ophthalmologic disorders, there is an unmet need for innovative diagnostic tools to support clinical decision-making. What did the study ask? This study evaluated the diagnostic performance of four AI models across three ophthalmologic subspecialties, testing the hypothesis that advanced language models can achieve accuracy levels comparable to human experts. What was learned from the study? Our results showed that ChatGPT-01 achieved the highest overall accuracy (84.9%), excelling in neuro-ophthalmology with a 95.5% accuracy, while DeepSeek and ChatGPT-4.0 each achieved 79.2%, and Qwens reached 64.2%. What specific outcomes were observed? In glaucoma, AI model accuracies ranged from 54.5% to 63.6%, with some models slightly surpassing the human expert average of 51.5%, underscoring the diagnostic difficulty of this condition. What has been learned and future implications? These findings highlight the potential of AI as a valuable adjunct to clinical judgment in ophthalmology, although further research and the integration of multimodal data are essential to optimize these tools for routine clinical practice.

Introduction The accurate prediction of mandibular bone growth is crucial in orthodontics and maxillofacial surgery, impacting treatment planning and patient outcomes. Traditional methods often fall short due to their reliance on linear models and clinician expertise, which are prone to human error and variability. Artificial intelligence (AI) and machine learning (ML) offer advanced alternatives, capable of processing complex datasets to provide more accurate predictions. This systematic review examines the efficacy of AI and ML models in predicting mandibular growth compared to traditional methods. Method. A systematic review was conducted following the PRISMA guidelines, focusing on studies published up to July 2024. Databases searched included PubMed, Embase, Scopus, and Web of Science. Studies were selected based on their use of AI and ML algorithms for predicting mandibular growth. A total of 31 studies were identified, with 6 meeting the inclusion criteria. Data were extracted on study characteristics, AI models used, and prediction accuracy. The risk of bias was assessed using the QUADAS-2 tool. Results. The review found that AI and ML models generally provided high accuracy in predicting mandibular growth. For instance, the LASSO model achieved an average error of 1.41 mm for predicting skeletal landmarks. However, not all AI models outperformed traditional methods; in some cases, deep learning models were less accurate than conventional growth prediction models. Discussion. The variability in datasets and study designs across the included studies posed challenges for comparing AI models’ effectiveness. Additionally, the complexity of AI models may limit their clinical applicability. Despite these challenges, AI and ML show significant promise in enhancing predictive accuracy for mandibular growth. Conclusion. AI and ML models have the potential to revolutionize mandibular growth prediction, offering greater accuracy and reliability than traditional methods. However, further research is needed to standardize methodologies, expand datasets, and improve model interpretability for clinical integration.

The article "Embracing Real AI: A Call to Action for Medical Physicists in Healthcare" urges medical physicists to prepare for the integration of artificial intelligence (AI) into healthcare practices, emphasizing their pivotal role in adapting to technological advancements. The authors advocate for embracing AI through advocacy, broadening perspectives, and enhancing coordination and communication. They propose an ABC strategy focusing on increasing educational initiatives, fostering interdisciplinary collaboration, and creating team collaboration to facilitate AI integration. The commentary highlights AI's potential in enhancing diagnostics, personalizing medicine, and automating routine tasks while addressing challenges such as data sharing and the role of federated learning. The article calls for medical physicists to lead in embracing AI, emphasizing continuous learning and collaboration to leverage its potential for improving healthcare and patient care. Medical physicists have consistently demonstrated strong interest in developing proficiency in the adoption of new technological advancements. The roots of the profession come from the radiation sciences, including radiation protection, radiation therapy, diagnostic imaging, and nuclear medicine.1 As science and technology continued to evolve, medical physicists' roles have extended into other non-radiation domains, such as non-ionizing-radiation-based imaging (ultrasound and magnetic resonance), molecular imaging, computer aided diagnosis (CAD), information technologies, and data science.2 In addition, medical physicists gradually have adopted increasingly more active roles in ensuring the professional education of other radiology/radiation oncology team members, maintaining high quality standards via quality assurance (QA) methods. They also play a major role in advising the hospital management on medical devices and software acquisition. The continuing expansion of these roles and responsibilities has put medical physicists on the forefront of embracing emerging technologies, making the profession one of the most technical and versatile in healthcare settings. Currently, as our field grows in importance, we medical physicists seek to continue to engage in significant ways to for increased contributions and roles in human health. This commentary/opinion urges medical physicists to prepare for their expanding roles in the field of AI and its implementation and oversight in clinical practice. Medical physicists must embrace "Real AI" to help integrate AI into healthcare practices. Conceptually we advocate for a strategy that involves Real AI through advocacy, broadening, and enhancing coordination/communication (an ABC strategy). In our current and future work medical physicists will use AI to automate routine tasks, allowing medical physicists to focus on more complex tasks. Furthermore, Medical Physics will use AI to enhance efficiency, safety, diagnostic and therapeutic applications, and for personalized medicine. However, as we have done in the past with other complex concepts (such as radiation), medical physicists need to be prepared for the potential risks and ethical dilemmas associated with AI, such as bias and lack of transparency. It will be important that Medical Physicists prepare for the rapidly changing AI landscape, and continue learning, gain hands-on experience, and collaborate with other AI experts in the healthcare environment. This paper aligns with the already approved guidance document developed by the AAPM in conjunction with International Atomic Energy Agency (IAEA)3 that discusses how medical physicists can ensure the effective implementation and management of AI systems. It is crucial for the Clinical Quality Management Program (CQMP) personnel to receive regular training and updates on relevant guidelines and legislation. Clear communication channels should be established with IT experts, vendors, and other stakeholders for smooth coordination.4 Comprehensive documentation should be developed to ensure compliance with contractual obligations and guidelines. The clinical team should be involved in acceptance testing and discussions, depending on the clinical purpose of the AI system.4 Protocols for data collection and curation should be established, along with the development of standardized validation datasets for performance evaluation.4 A system for monitoring updates to AI systems and models should be implemented, with the CQMP leading new acceptance/commissioning rounds for any updates. Lastly, mechanisms for continuous evaluation and improvement of the CQMP processes should be established, which could involve regular audits, feedback mechanisms from end-users, and incorporating lessons learned from previous rounds.4 Nowadays, major healthcare systems in the US consider their data as immensely valuable assets that require rigorous protection to ensure Health Insurance Portability and Accountability Act (HIPAA) compliance, as well as intellectual property considerations. It can be very difficult for researchers to share clinical data with vendors for development purposes without a significant return being specified to the institution, such as joint intellectual property or substantial grant funding. Instead, these healthcare systems encourage their researchers to commercialize their findings independently, allowing the institution to retain full rights to intellectual property. That said, the realization of federated learning would be a significant advancement. To achieve this, a powerful pre-trained model that would be adaptable to operation on different scales and in various clinical scenarios is necessary. It is plausible that local adaptation may not require substantial computing power or AI expertise. This concept is particularly intriguing and could be beneficial to smaller centers and clinics in underserved areas. However, the primary challenge is the cost. As we become more reliant on AI systems like OpenAI's ChatGPT or Google Gemini, we often overlook the fact that these conveniences come with a hefty price tag, costing billions of dollars to develop and maintain.5 As medical physicists we and other healthcare professionals can anticipate that AI will significantly transform healthcare, improving efficiency, accuracy, and the level of detail that can be extracted from imaging, and methods of therapy. These technological advancements are expected to bring immense value to the field, offering a new horizon in diagnostic and therapeutic capabilities. Yet, we also must recognize that it also introduces potential significant risks and ethical dilemmas. One of the primary concerns is the possibility of bias in AI, which can stem from the training data, the algorithms, or their application, leading to potentially detrimental effects on patient care. As medical physicists, we should acknowledge that the complexity and lack of transparency in AI decision-making processes present obstacles in terms of accountability and rectifying errors and requires greater oversight and responsibility. The integration of AI also has great capacity in redefining the role of medical physicists, impacting education and employment within the field. Addressing these issues necessitates the creation of ethical standards for AI in healthcare, emphasizing transparency, responsibility, and equity, with contributions from diverse stakeholders, including patients, medical professionals, and ethicists.6 Such measures are crucial to ensure the responsible utilization of AI in healthcare, and ultimately serve the best interests of patients and society. We anticipate that continued guidance from our professional societies will be helpful as our collective communities develop methods and approaches that help us learn, adopt, and employ AI responsibly. Advocacy: increase educational initiative, public awareness, and recommending processes at all levels of the clinical workforce, as well as patient engagement. Broadening Perspectives: encourage Interdisciplinary Collaborations that allow medical physicists to work with professionals from other disciplines such as computer science, data science, and biomedical engineering, to gain insights into different perspectives on AI applications in healthcare. This enables medical physicists to provide continuing education and connect the community with research opportunities. Improving Coordination and Communication through creating team collaboration: enhance communication with healthcare professionals, administrators, and patients by clearly defining and articulating the role of medical physicists in AI applications. Promote the sharing of knowledge, as exemplified by creating data repositories through contributions, to further creating the foundation of our understanding and application of AI in the field. We consider the concept of Real AI in our context to be aimed at providing and/or qualifying a ready AI product that has undergone a rigorous QA process, that is free of false additives and biases, with data carefully curated to represent the demographics and be attuned to the needs of the clinic, sourced with proper ingredients, and abiding by laws and regulations that can ensure the product serves the common health needs of patients and benefits the public's interest. What AI 'is' and what it 'is not' is a complex topic that warrants further exploration and understanding, but one vital for comprehension of what utility AI can fulfill in the clinical process, what its advantages and limitations are, and how it can be curated to perform in the clinical scenarios relevant to a particular radiology/radiation oncology practice. Multiple data-analysis algorithms have been created over the course of years, and not all of them qualify as AI.7 What distinction(s) lie in what constitutes AI? One possible interpretation is that AI is a system that can adapt to new data, or a system that generates insights driven by data. AI systems are designed to "learn" and adapt to new data and be stable over the course of introducing data perturbations or employ model adaptation mechanisms. AI systems can adjust the underlying data-processing mechanisms based on the input they receive, which allows them to improve their performance and make more accurate predictions or decisions over time. This is often achieved through techniques such as machine learning, where algorithms are trained on a dataset and then used to make predictions or decisions without being explicitly programed to perform the task.8 Understanding how such datasets are selected, what data needs to be fed into AI model to achieve desired results, and how to prevent common pitfalls and ethical conundrums associated with the use of AI models requires additional training that might yet be lacking in the traditional training of the radiology/radiation oncology adjacent specialists. The scope of involvement of each member of the team when it comes to AI integration into the clinic continues to be determined as the field rapidly evolves. When it comes to the role of medical physicists in conjunction with AI, an open discussion of the exact responsibilities is still ongoing, and feedback is encouraged from all the members of the community. So, what can medical physicists do? They can use AI to enhance quality improvement and safety by analyzing medical data to identify trends, patterns, and outliers.9 This can lead to the identification of areas for improvement or potential safety hazards and help them enter the realm of Responsible AI. AI can also improve diagnostic and therapeutic techniques by enhancing the quality of medical imaging and automating image interpretation.10 Furthermore, AI can help in integrating diagnostics, personalized medicine, and theragnostics by analyzing large datasets to tailor treatment plans to individual patients.11 This can lead to more effective and personalized care. AI can also automate routine tasks in medical physics, such as treatment planning and QA processes, leading to increased efficiency.12 Lastly, AI techniques like machine learning and deep learning can be leveraged for research and development to analyze complex datasets, discover patterns, and develop innovative techniques for disease detection, treatment, and monitoring.13 Whether it involves developing AI-driven solutions like automated segmentation, dose calculations, addressing intricate problems in the clinic, or potentially even contributing to open-source AI initiatives, such activities will empower medical physicists to enhance their skills and make tangible contributions to the advancement of healthcare. Embracing AI not only fosters a sense of accomplishment but also opens doors to the world of `automation' and scaling that will pervade all technologies of the future. The AHAIBC committee is at the center of bringing the medical physicist forward by developing curriculum concepts, bootcamps, and engendering engagement for our society. Integration of AI into the realm of medical physics education is critical, especially considering the potential significance of incorrect AI usage or misapplication. The physicist is responsible for installing and commissioning the AI software, ensuring the modeling is not biased, performing continuing QA on the hospital data and processes, and establishing efficient resource management. Embracing education in AI offers new benefits for medical physicists as it is already revolutionizing various industries and professional practices and we need to be equally prepared. One way to engage and prepare healthcare professionals for the upcoming AI wave is to start with the roots of quality safety and assurance. To do this, we should enable a comprehensive QA program that encompasses all clinical operations related to medical fields including radiology, nuclear medicine, and radiation oncology. Ensuring the safe operation of hardware, software, clinical operation processes and machinery is of utmost importance and one of the most crucial responsibilities of a medical physicist. A Real AI approach can be highly beneficial in achieving the goal of safe clinical implementation. Understanding the potential and limitations of AI serves as a cornerstone for fostering engagement not only within our profession but with other healthcare providers. Continuous learning and participation in hands-on experience are essential components for navigating the complexities of AI applications within healthcare. Collaboration, networking, and exploring AI's purpose and impact are equally vital in this journey. Additionally, some physicists may choose personal projects, embracing challenges in small groups, and actively contributing to AI-focused teams to amplify the motivation and expertise of our field. Insights through personal and collaborative opportunities ultimately provide for and encourage professional growth and innovation within our medical physics field. Some medical physicists may be able to attend specialty meetings and conferences dedicated to AI which further enriches their knowledge base and provides them avenues for fruitful collaboration. There are successful educational programs such as the Radiological Society of North America Artificial Intelligence (RSNA AI)-certificate program.14 Interdisciplinary cooperation and inter-institutional collaboration for AI experts is of paramount importance for integrating AI into medical physicists' practice on a larger scale, and mechanisms enabling this collaboration should be provided to the community. In summary, the authors believe that being prepared for and embracing the changes that AI is already bringing at the current time will benefit our community, healthcare, patient care, and society at large immediately and for the future. We are at a critical juncture, which can be considered a fourth industrial revolution, where AI and automation are applied more broadly. Medical physicists have a pivotal role to play in this revolution. We need to position ourselves at the forefront of 'Real AI' and lead the charge in this exciting new era. It is time for action, and we can take the first steps with potentially just a few ABCs. All authors contributed their efforts in writing and editing this call for action. ChatGPT search engine has been utilized to provide additional background to the subject of matter for illustrative purposes. The authors appreciate members of the Ad. The authors declare no conflicts of interest. The content for this call for action has been edited with the help of large language models ChatGPT and Google NotebookLM.

Artificial Intelligence: The Future of Maxillofacial Prognosis and Diagnosis?

Non-invasive and invasive imaging of vulnerable coronary plaque

e14162 Background: It is well established that an altered host metabolism has an impact on cancer outcome, possibly mediated by several mechanisms, including hyperglicaemia, hyperinsulinemia and presence of chronic inflammation. The aim of our analysis was to evaluate the correlation between host metabolism and clinical outcome in patients with advanced melanoma, kidney and non-small cell lung cancer (NSCLC), treated with immune checkpoint inhibitors (anti-CTLA4, anti PD1 and anti PDL1). Methods: The relationship between presence of type 2 diabetes mellitus (DMII) at baseline and outcome was assessed in 187 patients treated with immune checkpoint inhibitors in two cancer centers. Progression Free Survival (PFS) and Overall Survival (OS) were calculated by Kaplan-Meier estimation; multivariate Cox analysis was performed according to age, gender, BMI (normal < 25 kg/m2, overweight 25-30 kg/m2, obese > 30 kg/m2), type of cancer and line of treatment. Results: One-hundred-sixty-eight patients were available for our analysis. Twenty-eight patients (17%) were diabetic at baseline. Median age was 65 (range 25-80); 83 patients were males (49%); 82 (48%) had advanced melanoma, 83 (49%) NSCLC and 3 (3%) kidney cancer. One-hundred-two (60%) patients had BMI < 25, 51 (30%) were overweight and 16 (10%) were obese. The first line of treatment was immunotherapy in 83 (49%) patients. By univariable analysis median PFS was 4.2 months in non diabetics vs 6.4 in diabetics patients (HR 0.95; 95%CI 0.58-1.58); median OS was 6.17 and 9.1 months, respectively (HR 1.00; 95%CI 0.58-1.75). At multivariable analysis, taking into account DMII, BMI, sex, age, line of treatment and type of cancer, we found that BMI ≤25 was associated with a two fold increase in risk of progression (PD) or death (p = 0.005), whereas patients who received immunotherapy as second or subsequent line had a two fold increase in risk of PD or death (p = 0.003). Conclusions: The results of our analysis show that in patients with advanced cancer treated with immune checkpoint inhibitors, the presence of DMII does not adversely affect the clinical outcome. Conversely, lower BMI was associated with a significantly worse PFS and OS, independently from type of cancer, age and gender. As expected, patients who received immunotherapy in later lines of treatment had a significantly shorter survival.

Background The choice of treatment for inflammatory bowel disease (IBD) with multiple prior drug failures, also known as difficult-to-treat (D2T) IBD, is an increasing challenge. We assessed the persistence of biologics and advanced small molecules, a surrogate for their efficacy, in patients with Crohn's disease (CD) and ulcerative colitis (UC) overall and in 4th, 5th and 6th lines of advanced treatment. Methods We retrospectively searched the electronic medical records of patients with IBD followed at the San Raffaele Hospital (Milan, Italy) up to 1 October 2023. Patients enrolled in clinical trials that changed disease management or receiving experimental drugs were excluded. Drug persistence was defined as the time from initiation to discontinuation of each treatment. Welch's ANOVA and Games-Howell's method for multiple comparisons were used. Results A total of 679 patients with moderate to severe IBD were included. 350 had CD, 326 UC, and 3 IBD-U. The mean disease duration was 11 years and 452 (66%) had received 5-aminosalicylates or antimetabolites as first maintenance treatment. Escalation to and changes between advanced treatments over time is summarised in Figure 1A. A greater use of anti-TNF agents was observed in the first lines of treatment whereas newer medications were proportionally more prescribed in refractory patients. In the overall analysis, regardless of the line of treatment, there was no difference in drug persistence between agents at 12 months (p=0.62), with approximately half of the patients having discontinued the drug. Figure 1B In subsequent lines of treatment, drug persistence decreased significantly in patients with CD (p=0.002). The trend was particularly pronounced from line 2, with a mean duration of 31 months, to line 7, with a mean duration of 11 months. In UC, the reduction was less evident and borderline non-significant (p=0.05), although it suggests a decrease in drug efficacy as the disease progresses. Figure 2 A, B In patients with D2T IBD, drug persistence in 4th, 5th and 6th line advanced treatment was similar for all agents (all p>0.05), indicating no clear advantage of one drug over the others. Figure 2 C, D, E Conclusion In CD, and possibly in UC, the persistence of patients on any advanced drug decreases with subsequent lines of treatment. None of the advanced agents showed a longer persistence time when used as a 4th, 5th or 6th line of treatment.

BackgroundQualitative research appraisal is crucial for ensuring credible findings but faces challenges due to human variability. Artificial intelligence (AI) models have the potential to enhance the efficiency and consistency of qualitative research assessments.ObjectiveThis study aims to evaluate the performance of 5 AI models (GPT-3.5, Claude 3.5, Sonar Huge, GPT-4, and Claude 3 Opus) in assessing the quality of qualitative research using 3 standardized tools: Critical Appraisal Skills Programme (CASP), Joanna Briggs Institute (JBI) checklist, and Evaluative Tools for Qualitative Studies (ETQS).MethodsAI-generated assessments of 3 peer-reviewed qualitative papers in health and physical activity–related research were analyzed. The study examined systematic affirmation bias, interrater reliability, and tool-dependent disagreements across the AI models. Sensitivity analysis was conducted to evaluate the impact of excluding specific models on agreement levels.ResultsResults revealed a systematic affirmation bias across all AI models, with “Yes” rates ranging from 75.9% (145/191; Claude 3 Opus) to 85.4% (164/192; Claude 3.5). GPT-4 diverged significantly, showing lower agreement (“Yes”: 115/192, 59.9%) and higher uncertainty (“Cannot tell”: 69/192, 35.9%). Proprietary models (GPT-3.5 and Claude 3.5) demonstrated near-perfect alignment (Cramer V=0.891; P<.001), while open-source models showed greater variability. Interrater reliability varied by assessment tool, with CASP achieving the highest baseline consensus (Krippendorff α=0.653), followed by JBI (α=0.477), and ETQS scoring lowest (α=0.376). Sensitivity analysis revealed that excluding GPT-4 increased CASP agreement by 20% (α=0.784), while removing Sonar Huge improved JBI agreement by 18% (α=0.561). ETQS showed marginal improvements when excluding GPT-4 or Claude 3 Opus (+9%, α=0.409). Tool-dependent disagreements were evident, particularly in ETQS criteria, highlighting AI’s current limitations in contextual interpretation.ConclusionsThe findings demonstrate that AI models exhibit both promise and limitations as evaluators of qualitative research quality. While they enhance efficiency, AI models struggle with reaching consensus in areas requiring nuanced interpretation, particularly for contextual criteria. The study underscores the importance of hybrid frameworks that integrate AI scalability with human oversight, especially for contextual judgment. Future research should prioritize developing AI training protocols that emphasize qualitative epistemology, benchmarking AI performance against expert panels to validate accuracy thresholds, and establishing ethical guidelines for disclosing AI’s role in systematic reviews. As qualitative methodologies evolve alongside AI capabilities, the path forward lies in collaborative human-AI workflows that leverage AI’s efficiency while preserving human expertise for interpretive tasks.