Citations

Abstract Radiological protocol selection is a critical but time-consuming step in clinical workflow, requiring radiologists to match patient indications with the appropriate MRI or CT protocol. Manual selection can be prone to delays or potential errors, and automated approaches must contend with substantial data imbalance, site-specific variation, and evolving nomenclature. We investigated whether a large language model (LLM) can support reliable protocol selection at scale and whether retrieval-augmented generation (RAG) offers operational advantages over direct fine-tuning. Using 498,228 patient reports collected across three Mayo Clinic sites (Arizona, Florida, and Rochester) spanning six radiological divisions, we trained site-specific Llama 3.2 3B models for use with and without retrieval augmentation. Division-scoped Facebook AI Similarity Search (FAISS) indexes constructed from procedure and diagnosis text were used to supply contextual evidence in the RAG framework. Both fine-tuned and RAG-augmented models achieved strong performance across sites, with F1 scores of 0.88–0.90. RAG matched or modestly trailed direct fine-tuning overall but delivered consistent gains in specific divisions (e.g., musculoskeletal imaging). Importantly, the RAG model introduced abstention behavior (Not Enough Information), which concentrated in linguistically diverse divisions and provided an interpretable signal of uncertainty. These findings suggest that RAG-based models are viable for division-scoped protocol selection and offer practical advantages. Retrieval indexes can be refreshed far more easily and with fewer resources than retraining LLMs, enabling continual adaptation to evolving clinical workflows. Future prospective deployment will evaluate real-time accuracy, agreement with practitioners, and the role of abstention as a safety mechanism in clinical decision support.

ABSTRACT Objective Autosomal dominant progranulin ( GRN ) pathogenic variants are a genetic cause of frontotemporal lobar degeneration. Though clinical trials for GRN ‐related therapies are underway, there is an unmet need for biomarkers that can predict symptom onset and track disease progression. We previously showed that asymptomatic GRN variant carriers exhibit thalamocortical hyperconnectivity that increases with age, presumably as they are approaching symptom onset. Whether hyperconnectivity arises concomitantly with markers of neurodegeneration remains unclear. Methods Utilizing T1 and task‐free functional magnetic resonance imaging from 49 asymptomatic and 26 symptomatic GRN variant carriers, we determined the relationships between functional connectivity, as measured by voxelwise whole‐brain degree, and GRN ‐relevant markers of disease progression, including plasma neurofilament light chain concentrations, cerebrospinal fluid complement C1q and C3b protein levels, obsessive‐compulsive disorder symptom severity, and gray matter volume. Results Neurofilament light chain concentrations were associated with frontotemporoparietal and thalamic hyperconnectivity in asymptomatic GRN variant carriers and extensive regions of atrophy in symptomatic carriers. Complement levels were associated with regions of hyperconnectivity, but not gray matter volume, in symptomatic carriers. Obsessive‐compulsive disorder symptom severity was associated with hypoconnectivity across all GRN carriers. Asymptomatic carriers with thalamic hyperconnectivity tended to have lower gray matter volume in the bilateral insula and left lateral parietal cortex, early regions of atrophy in GRN ‐frontotemporal dementia. Interpretation In asymptomatic carriers, the co‐occurrence of hyperconnectivity, high neurofilament light chain, and low gray matter volume suggests that functional hyperconnectivity may portend the onset of clinical decline. These findings point toward hyperconnectivity as an indicator of approaching symptomatic onset.

Abstract Ovarian cancer is the deadliest gynecologic malignancy worldwide, with a 5-year overall survival rate of approximately 49%. Although complete resection is associated with the most favorable prognosis following primary debulking surgery, accurately assessing tumor resectability at diagnosis remains a major clinical challenge. We propose a decision support system (DSS) designed to predict residual tumor after primary debulking surgery, based on clinical and imaging data available at diagnosis. The system was developed and validated using a retrospective cohort of 465 patients with high-grade serous ovarian cancer, collected at the European Institute of Oncology in Milan, Italy. We developed a deep learning (DL) model that combines pre-trained Vision Transformer encoders with an attention mechanism and a classification head. When evaluated on an independent test cohort of 75 cases, our best-performing model achieved an area under the curve (AUC) of 0.80 (95% CI: 0.68–0.89) and a recall of 0.86 (95% CI: 0.71–0.97), demonstrating discriminatory ability with a particularly low rate of false negatives. Clinically, the model correctly identified 24 out of 28 patients (86%) with non-resectable disease, who would not have benefited from primary debulking surgery. Notably, within this subgroup, the model accurately predicted 93% (13 out of 14) of cases in which surgery was aborted intraoperatively due to unforeseen unresectable disease. These findings suggest that the model could have potential in preventing unnecessary and inappropriate surgical interventions. The proposed DSS is a fully automated, DL–based system for predicting tumor resectability at diagnosis, without the need for manual segmentation or expert evaluation of radiological features and structured clinical parameters. This approach could facilitate and accelerate personalized treatment planning in ovarian cancer. The work is part of the project Under-XAI: understanding ovarian cancer initiation and progression through explainable AI. Project code: PNRR-MAD-2022-12376574.

Technological innovations have improved many barriers in lung transplantation, but high rates of acute and chronic rejection still limit lung allograft survival. This may be explained by the unique environment of the lung. As a mucosal barrier organ, the lung is constantly exposed to the external environment, leading to unique immunological features that are not seen in other transplantable solid organ allografts such as hearts, kidneys, and livers. Thus, the higher rates of rejection and poor long-term survival of lung transplant recipients may stem from the global immunosuppression strategies that are indiscriminately used for all solid organ grafts. Data from our laboratory, and others, have demonstrated that the unique immunoregulatory pathways of the lung may require different strategies for long-term graft survival. For example, depletion of CD8 + T cells typically contributes to the acceptance of transplanted organs. However, these cells facilitate lung allograft acceptance through interferon gamma mediated nitric oxide production. Interestingly CD8 + T cells modify and polarize eosinophils to produce nitric oxide as a means of tolerance induction. Such eosinophils also function to maintain long-term lung allograft acceptance by interfering with humoral alloimmunity. For most other organs eosinophils are suspected to contribute to graft rejection. In this review, we aim to describe the cytokine pathways involved in lung allograft rejection or tolerance, contrast such pathways to those evident in other solid organs, and discuss the need for further studies that can be used to design rational methods for altering the cytokine environment to improve lung allograft survival.

ABSTRACT Background Ventricular tachycardia (VT) is a major contributor to sudden cardiac death among patients with ischemic heart disease (IHD). While catheter ablation reduces the arrhythmic burden, the impact on long‐term outcomes remains uncertain. We assess the efficacy of catheter ablation in patients with IHD. Methods We performed a meta‐analysis by systematically searching electronic databases from inception to April 2025. Pairwise, reconstructed time‐to‐event data meta‐analyses and Trial Sequential analyses were conducted. Results A total of seven studies (1192 participants) were included. Among patients with IHD with ICD, VT ablation was associated with a significant reduction in ICD shocks (RR = 0.50, 95% CI = [0.34–0.74]), a trend toward reduced VT storm (RR = 0.64, 95% CI = [0.40–1.04], p = 0.07), and lower rates of cardiovascular hospitalization (RR = 0.73, 95% CI = [0.53–1.01], p = 0.06) compared to ICD alone. However, no significant differences were observed when compared to patients receiving ICD plus antiarrhythmic drugs (AADs). VT ablation showed no significant impact on mortality or VT/VF recurrence compared to either ICD alone or ICD with AADs. Reconstructed Kaplan–Meier analysis for all‐cause mortality showed no significant difference between VT ablation and AAD + ICD. Trial Sequential Analysis provided conclusive evidence for VT/VF recurrence, while further data is required for other outcomes. Conclusion Among patients with IHD with ICD, VT ablation reduces ICD shocks, VT storm, and cardiovascular hospitalizations compared to ICD‐only therapy, but offers no significant advantage over ICD combined with AADs. VT ablation does not impact overall mortality or the recurrence of VT/VF. Trial sequential analysis confirmed conclusive evidence for VT/VF recurrence, while additional data is needed for other outcomes.

AbstractTo review research on the effectiveness of computer-based auditory training (CBAT) in new adult cochlear implant (CI) recipients.Studies by the primary author on auditory training in new adult CI recipients were reviewed, focusing on a database of auditory training data in >100 patients followed over their first-year post-activation.CBAT, as used in a standard outpatient setting, affords benefits compared to other commonly employed exercises/programs. In new adult CI users, CBAT was associated with greater improvement in CNC words, AzBio sentences, and Cochlear Implant Quality of Life-35 (CIQOL-35) Profile scores at 3 months post-activation compared to CI recipients who did not train or used other resources. By 12 months, patients showed similar improvement in speech recognition scores, regardless of training modality; however, patients who used CBAT continued to show significantly greater improvement in CIQOL global and all domain scores compared to CBAT nonusers. This benefit was noted regardless of whether CBAT was started before or after 3 months post-activation. No demographic/lifestyle factors, audiological measures, or pre-CI CIQOL/CIQOL-Expectations scores were associated with CBAT use.CBAT during the first-year post-activation may contribute to early gains in CI speech recognition and to persistent improvements in functional abilities, as measured via the CIQOL-35 profile, in new implant recipients. Consideration of specific recommendations for the use of CBAT resources may be valuable, given the minimal risk, ease of access, and affordability of these programs.

Objective: To evaluate the documentation of menopause-related International Classification of Diseases-10 (ICD-10) codes in the electronic health record (EHR) among midlife women with moderate or greater menopause symptoms receiving primary care. Methods: This cross-sectional study from the Hormones and Experiences of Aging (HERA) cohort included women aged 45-60 years receiving primary care at one of 4 Mayo Clinic sites who completed a one-time survey between March 1, 2021 and June 30, 2021. The survey captured demographic data, menopause symptoms using the Menopause Rating Scale (MRS), health care utilization, and treatment. Women with an MRS score ≥12 were included. The primary outcome was documentation of a menopause-related ICD-10 code in the EHR in the 12 months before survey completion. Results: Of 5,254 women with completed surveys, 2,414 (49%) had an MRS score ≥12 and were included. Among these, 1,519 (63%) reported seeking care for their menopause symptoms in the past 12 months, but only 345 (23%) had a menopause-related ICD-10 code. Women with an ICD-10 code had higher MRS scores (18 [IQR: 14-22] vs 17 [IQR: 14-20]; P = 0.002) and were more likely to use systemic hormone therapy (HT; 26% vs 9%; P < 0.001), and vaginal HT (20% vs 6%; P < 0.001). Conclusions: Menopause-related ICD-10 diagnosis codes were under-documented in women with significant menopause symptom burden. Those with a code were more likely to report menopause treatment. These findings highlight a critical gap between symptom burden and diagnosis coding in the EHR, underscoring the need to improve identification and management of menopause symptoms.

PURPOSE Preclinical studies suggest the combination of the poly(adenosine diphosphate[ADP]-ribose) polymerase (PARP) inhibitor olaparib and the ataxia telangiectasia–mutated and Rad3-related (ATR) inhibitor ceralasertib has synergistic antitumor activity in homologous recombination proficient (HRP) and homologous recombination repair gene mutation (HRRm) patients. This study aims to determine the efficacy of olaparib plus ceralasertib in metastatic castration-resistant prostate cancer (mCRPC) with and without HRRm. PATIENTS AND METHODS Thirty-five HRP and 12 HRRm PARP inhibitor–naïve patients with mCRPC progressing on ≥1 line of therapy or previous androgen receptor pathway inhibitor were enrolled. Patients received olaparib (300 mg twice daily) and ceralasertib (160 mg once daily, days 1-7) in 28-day cycles. HRRm was defined as biallelic inactivation of BRCA2 or BRCA1 , monoallelic inactivation of ATM , or germline inactivation of any of these genes. The primary end point was disease response rate (dRR; confirmed prostate-specific antigen decline ≥50% and/or complete/partial radiographic response by RECIST v1.1) in HRP patients. Secondary end points included dRR in HRRm patients, progression-free survival (PFS), and safety/toxicity. Two-stage designs were used. RESULTS Four of 35 (11%; 95% CI, 3.2 to 26.7) HRP and four of 12 HRRm patients responded. Median PFS was 8.2 months (95% CI, 5.3 to not reached) for HRP patients. Thirty-six percent of patients had ≥grade 3 toxicity, most commonly from anemia. Limitations include small, single-arm, nonrandomized trial design. CONCLUSION Combining ceralasertib with olaparib had limited activity in patients with HRP mCRPC. HRRm response rate was not greater than previous single-agent PARP inhibitor clinical trials.