Citations

The integration of artificial intelligence (AI) in medical imaging has revolutionized diagnostic capabilities, yet the black-box nature of deep learning models poses significant challenges for clinical adoption. Current explainable AI (XAI) approaches, including SHAP, LIME, and Grad-CAM, predominantly focus on post hoc explanations that may inadvertently undermine clinical decision-making by providing misleading confidence in AI outputs. This paper presents a systematic review and meta-analysis of 67 studies (covering 23 radiology, 19 pathology, and 25 ophthalmology applications) evaluating XAI fidelity, stability, and performance trade-offs across medical imaging modalities. Our meta-analysis of 847 initially identified studies reveals that LIME achieves superior fidelity (0.81, 95% CI: 0.78–0.84) compared to SHAP (0.38, 95% CI: 0.35–0.41) and Grad-CAM (0.54, 95% CI: 0.51–0.57) across all modalities. Post hoc explanations demonstrated poor stability under noise perturbation, with SHAP showing 53% degradation in ophthalmology applications (ρ = 0.42 at 10% noise) compared to 11% in radiology (ρ = 0.89). We demonstrate a consistent 5–7% AUC performance penalty for interpretable models but identify modality-specific stability patterns suggesting that tailored XAI approaches are necessary. Based on these empirical findings, we propose a comprehensive three-pillar accountability framework that prioritizes transparency in model development, interpretability in architecture design, and a cautious deployment of post hoc explanations with explicit uncertainty quantification. This approach offers a pathway toward genuinely accountable AI systems that enhance rather than compromise clinical decision-making quality and patient safety.

Abstract Background Dose‐driven continuous scanning (DDCS) is a type of beam delivery in Hitachi pencil beam scanning (PBS) particle therapy system. Particle beam is irradiated continuously between spots to achieve a higher effective dose rate and a shorter beam delivery time. The beam is only turned off at break spots, the end of a treatment layer and during spill change. The break spot is a unique feature of the DDCS introduced by the treatment planning system (TPS) to interrupt continuous scanning. DDCS separates the dose delivery of a spot into a move dose and stop dose components. The move dose is controlled by the scanning speed and beam current, while the stop dose delivers the remaining monitor units (MUs). Hence, rigorous quality assurance (QA) of the scanning speed and functionality of break spot are required to ensure safe and successful DDCS delivery. Purpose We aim to report for the first time the QA methodology and result for DDCS, focusing on the scanning speed and break spot validation. Methods A rectangular spiral irradiation pattern with different degree of break spots were irradiated on a 2D strip ionization chamber (CROSSmini) detector using 50 time resolution to validate the break spots. Four methods of measuring scanning speeds were proposed. Using a spiral irradiation pattern, methods one and two applied linear regression to the scanning time measured by the CROSSmini for different inter‐spot distances, with measurements taken at sampling rates of 5 and 20 kHz, respectively. Methods three and four measured the scanning time of single inter‐spot distances in the X and Y directions. Method three measured the scanning time with the CROSSmini while method four measured the rise and fall time of the scanning magnet (SCM) current. All measurements were performed with 70.2, 150.2 and 228.7 MeV proton beam energies and three different beam currents of 8, 14 and 20 MU/s. Lastly, the scanning speed tolerance limit was calculated analytically under the requirement of avoiding a beam abort when the stop dose was less than zero. Results The data recorded by the CROSSmini under the spiral irradiation patterns with break spots was reconstructed into a 2D dose, and the break spots’ operations were visually validated for all the energies and beam currents. There were no statistically significant difference in the measured X and Y scanning speeds across the four methods. However, the last method involving the SCM current yielded the largest type A uncertainty of up to 25% () due to the noisy signal. The first method with the linear regression and 5 kHz acquisition rate was preferred due to a smaller type A uncertainty of up to 2.0% and a lower false negative rate in detecting the rising and falling edges during scanning. The calculated tolerance limit based on our current DDCS setting was found to be 0.50 times of the expected scanning speed. Conclusion We have introduced QA protocols for DDCS, designed to ensure complete and safe beam delivery without interruptions caused by move doses exceeding planned spot doses.

Effective healthcare referral triage poses a significant challenge, especially extracting referral notes on time for scheduling. To offset this issue, this study aims to develop an optimized Random Forest classifier to recreate the referral notes content and identify patients requiring Spinal Cord Stimulation procedures. A performance index was introduced to improve accuracy, while a consistency index was designed to prevent overfitting. The optimized classifier, with the optimal selection at the number of estimates being 5 trees and tree depth being 4, achieved an accuracy of 0.74 with a lower standard deviation, demonstrating its reliability and effectiveness.

ABSTRACT Osteoporosis is a chronic disease mainly prevalent in older people. A hip fracture, its most devastating complication, is associated with a one in five chance of mortality in the first year. This review discusses emerging risk factors for bone fragility including diabetes, sarcopenia, and bariatric surgery. Recency of fracture represents an imminent risk that deserves prompt recognition. The new FRAXplus calculator accounts for additional risk factors to refine existing risk stratification. Advances in imaging techniques such as trabecular bone score and radiofrequency echographic multi-spectrometry may have potential roles in specific populations. Treatment of osteoporosis requires a multifaceted approach. Non-drug interventions include fall prevention and resistance exercises, supplementation with calcium, vitamin D and possibly vitamin K, and effective bridging of care gaps with fracture liaison services. An overview of drug therapy is provided, drawing attention to the importance of sequential therapy starting with potent anabolic agents for the patients at the highest risk, as opposed to a traditional step-up approach. An individualized, goal directed care plan is necessary for a treat-to-target approach, with total hip bone mineral density being the best specific target. Future research in cellular senescence points to the role of senolytics and senomorphics to alleviate osteoporosis, among other age related conditions.

Abstract INTRODUCTION African American (AA) individuals are underrepresented in biomarker studies for Alzheimer's disease (AD). Biomarkers that reflect the heterogeneity of AD and achieve the greatest accuracy across populations are sorely needed. METHODS Untargeted proteome measurements were obtained using the SomaScan 7k platform to identify novel plasma biomarkers for AD in AA participants with clinical diagnoses of AD dementia ( n = 181) and cognitively unimpaired (CU, n = 142). Machine learning was used to identify a set of plasma proteins that yielded the best classification accuracy. RESULTS A set of 36 proteins achieved an area under the curve (AUC) of 0.94 to classify AD dementia versus CU, a 16% improvement over age, sex, and apolipoprotein E ( APOE) . This finding was replicated in multiple plasma and brain datasets (AUCs 0.73–0.97). Our findings underscore the importance of matrisome and cerebrovascular dysfunction in AD pathophysiology. DISCUSSION This study demonstrates the potential of biomarker discovery through untargeted plasma proteomics and machine learning. Highlights Conducted large‐scale plasma proteomics in Alzheimer's disease (AD) versus cognitively unimpaired controls. Machine learning biomarker discovery was replicated in an independent cohort. Novel set of proteins distinguishes AD versus controls with high accuracy (area under the curve [AUC] = 0.94). Achieved reproducibility across multiple replication cohorts (AUC = 0.73–0.97). Network analyses implicates matrisome biology and cerebrovascular dysfunction.

ABSTRACT Introduction Although liver transplantation (LT) rates have increased globally, the demand for organs still exceeds the supply. This review highlights cultures with significant influences on organ donation and LT. Methods We performed a literature search for articles related to cultures, ethnologies, philosophies, policies, religions, socio‐economic factors, and other influences affecting organ donation in LT around the world. Results There are astonishing similarities between different cultures. In Africa, Asia, and Europe, people often view the postmortem bodies as sacred and believe that their bodies are gifts from “God” or ancestors. To respect the gift of life, to have a chance in reincarnation, or to enjoy life after death, many believe that their bodily sanctity must be preserved. In Asia, Europe, and Latin America, there is a general reluctance to speak about death. Religions also play an important role in people's views on organ donation. Different regions around the world have similar challenges to organ donation, such as misconceptions about organ trafficking, political and economic instability, and a lack of governmental or religious support. Conclusion Addressing the complexities of organ donation and LT requires a multifaceted approach. Culturally sensitive education tailored to diverse communities can dispel myths and misconceptions surrounding organ donation. Fostering partnerships between healthcare providers, religious leaders, community organizations, and policymakers can further facilitate open dialogues and promote organ donation as a noble act to save lives. Social Media Similar cultures around the world post challenges for organ donation. Organizations should foster partnerships between healthcare providers, politicians, and religious leaders to promote donation.

Background/Objectives: Proton therapy delivers highly conformal doses to the target area without producing an exit dose, minimizing cumulative doses to healthy liver tissue. This study aims to evaluate current practices, challenges, and variations in the implementation of proton stereotactic body radiation therapy (SBRT) and hypofractionated therapy for liver malignancies, with the goal of providing a technical assessment to promote broader adoption and support future clinical trials. Methods and Materials: An extensive survey was conducted by NRG Oncology across North American proton treatment centers to assess the current practices of proton liver SBRT and hypofractionated therapy. The survey focused on key aspects, including patient selection, prescription and normal tissue constraints, simulation and motion management, treatment planning, quality assurance (QA), treatment delivery, and the use of image-guided radiation therapy (IGRT). Results: This survey captures the current practice patterns and status of proton SBRT and hypofractionated therapy in liver cancer treatment. Proton therapy is increasingly preferred for treating inoperable liver malignancies due to its ability to minimize healthy tissue exposure. However, the precision required for proton therapy presents challenges, particularly in managing uncertainties and target motion during high-dose fractions and short treatment courses. Survey findings revealed significant variability in clinical practices across centers, highlighting differences in motion management, dose fractionation schedules, and QA protocols. Conclusion: Proton SBRT and hypofractionated therapy offer significant potential for treating liver malignancies. A comprehensive approach involving precise patient selection, treatment planning, and QA is essential for ensuring safety and effectiveness. This survey provides valuable insights into current practices and challenges, offering a foundation for technical recommendations to optimize the use of proton therapy and guide future clinical trials.

Abstract Background The radiation oncology clinical practice involves many steps relying on the dynamic interplay of abundant text data. Large language models have displayed remarkable capabilities in processing complex text information. But their direct applications in specific fields like radiation oncology remain underexplored. Purpose This study aims to investigate whether fine‐tuning LLMs with domain knowledge can improve the performance on Task (1) treatment regimen generation, Task (2) treatment modality selection (photon, proton, electron, or brachytherapy), and Task (3) ICD‐10 code prediction in radiation oncology. Methods Data for 15 724 patient cases were extracted. Cases where patients had a single diagnostic record, and a clearly identifiable primary treatment plan were selected for preprocessing and manual annotation to have 7903 cases of the patient diagnosis, treatment plan, treatment modality, and ICD‐10 code. Each case was used to construct a pair consisting of patient diagnostics details and an answer (treatment regimen, treatment modality, or ICD‐10 code, respectively) for the supervised fine‐tuning of these three tasks. Open source LLaMA2‐7B and Mistral‐7B models were utilized for the fine‐tuning with the Low‐Rank Approximations method. Accuracy and ROUGE‐1 score were reported for the fine‐tuned models and original models. Clinical evaluation was performed on Task (1) by radiation oncologists, while precision, recall, and F‐1 score were evaluated for Task (2) and (3). One‐sided Wilcoxon signed‐rank tests were used to statistically analyze the results. Results Fine‐tuned LLMs outperformed original LLMs across all tasks with p value ≤ 0.001. Clinical evaluation demonstrated that over 60% of the fine‐tuned LLMs‐generated treatment regimens were clinically acceptable. Precision, recall, and F1‐score showed improved performance of fine‐tuned LLMs. Conclusion Fine‐tuned LLMs demonstrated statistically significant improvements over original LLMs upon three clinically important tasks in radiation oncology. This study explored the feasibility of applying fine‐tuned LLMs in radiation oncology, inspiring further development of utilizing LLMs to assist with radiation oncology tasks.

Abstract Teclistamab is an anti-CD3-/B-cell maturation antigen (BCMA) bispecific antibody approved for use in relapsed/refractory multiple myeloma. We undertook a retrospective study of post-approval, real-world outcomes with teclistamab in the US Multiple Myeloma Immunotherapy Consortium. Among 509 patients, 89% would have been ineligible for the MajesTEC-1 trial, primarily due to prior BCMA-directed therapy, cytopenias, or diminished performance status. Cytokine release syndrome occurred in 54% (1.4% grade ≥3) and immune effector cell–associated neurotoxicity syndrome in 11% (2.2% grade ≥3), with no fatal events. Infections occurred in 42% and contributed to death in 5%. Partial response (PR) or better was achieved in 53% and very good PR or better in 45%. With 10.1 months of median follow-up, the estimated median progression-free survival (PFS) was 5.8 months and 12-month overall survival was 61%. Independent predictors of less than very good PR and shorter PFS included BCMA-directed chimeric antigen receptor T-cell therapy in the previous 9 months, high disease burden, lymphopenia, and elevated ferritin. Significance: T cell–engaging bispecific antibodies such as teclistamab represent an important new treatment modality for multiple myeloma and other blood cancers. This study evaluates the real-world safety and efficacy of teclistamab, including its activity in populations not represented in the initial phase I/II study, and identifies clinical variables associated with treatment response. See related commentary by Zweegman et al., p. 542