Citations

Abstract Immune checkpoint inhibitors (ICIs) can produce durable tumor control, yet many patients experience primary or acquired resistance, mixed lesion-level responses, or immune-related adverse events. Tissue and blood biomarkers provide biologically informative but spatially incomplete views of a heterogeneous and time-varying tumor–host system. Radiologic imaging complements these assays by repeatedly measuring whole-body tumor burden, organ-specific disease patterns, lesion-level heterogeneity, and treatment-related change. This critical narrative review evaluates imaging-centered artificial intelligence (AI) for precision cancer immunotherapy, including baseline and serial CT, MRI, and PET; radiomics and learned image representations; lesion-level and whole-body heterogeneity; toxicity assessment; and integration with pathology, genomics, circulating tumor DNA, laboratory measurements, and electronic health records. Representative cohorts range from early radiogenomic studies of approximately 100–200 patients to externally validated and longitudinal models involving hundreds or thousands of patients. Rather than treating all reported performance estimates as equally actionable, we appraise the clinical question, cohort design, endpoint, validation strategy, calibration, and intended clinical action. Current evidence supports noninvasive immune phenotyping, risk stratification, longitudinal monitoring, and multimodal biomarker integration. However, most models predict outcomes among patients already receiving ICIs and therefore remain prognostic under treatment rather than predictive of differential treatment benefit. Radiotherapy (RT) combined with ICI serves as a focused imaging use case because lesion identity, dose distribution, fractionation, sequence, and thoracic and normal-tissue exposure must be linked to defined actions. Existing patient-level AI studies estimate efficacy or toxicity after treatment selection; none has prospectively shown that an AI-selected RT strategy improves patient outcomes. Translation will require patient-level leakage control, calibration, geographic and temporal validation, robustness to missing modalities, and prospective evaluation of prespecified model-guided decisions.

Abstract Objective Genicular nerve radiofrequency (gRF) treatment is a widely used intervention for chronic pain from knee osteoarthritis (KOA). This systematic review and meta-analysis focuses on RCTs utilizing image guidance to address inconsistencies in previous reviews and provide higher-quality evidence. Design, Setting, and Methods This study included RCTs that compared gRF techniques (conventional, cooled, pulsed, bipolar) to non-gRF treatments, including sham procedures, intra-articular injections, ozone therapy, and oral medications. Primary and secondary outcomes were changes in pain and total WOMAC scores at 3, 6, and 12 months post-gRF treatments. Subgroup analyses explored differences by lesion size, use of prognostic blocks, industry sponsorship, and study location. Statistical heterogeneity was assessed via I 2 statistics. Results Eleven RCTs with 868 knee osteoarthritis patients (435 gRF, 433 non-gRF) were included. At 3 months, gRF significantly reduced pain (SMD -1.65, 95% CI: -3.20 to -0.10, I 2 = 98%, P = 0.040), but not at 6 months or 12 months. Secondary analyses showed no significant differences in total WOMAC scores between the gRF and control groups at 3, 6, and 12 months. Subgroup analysis showed that industry-sponsored studies reported significant 3-month pain reduction, whereas non-sponsored studies did not. This effect was independent of study location. Conclusions GRF provides short-term pain relief (up to 3 months) but neither functional improvement nor sustained pain reduction compared to non-gRF treatments. Our findings do not support the routine use of gRF for long-term chronic pain due to KOA.

Retrieval-augmented generation (RAG) is rapidly emerging as a transformative paradigm for large language models (LLMs), especially in high-stakes domains like oncology that demand precision, factual grounding, and up-to-date knowledge. By pairing LLMs with external knowledge repositories, RAG systems explicitly ground model outputs in relevant retrieved documents, helping to reduce hallucinations and ensure responses reflect current evidence. In oncology, where clinical knowledge evolves continually with new research and drug approvals, RAG offers a way to integrate the latest data (e.g., trial results, guidelines, genomic databases) into decision-making. This review synthesizes the technical foundations of RAG, including its architecture and key components, and examines current applications in oncology such as clinical decision support, patient education, radiology reporting, pathology analysis, and genomics-driven precision medicine. We highlight recent studies that demonstrate RAG’s potential—for instance, improving treatment recommendations by incorporating genetic profiles and literature, and enhancing diagnostic accuracy by integrating guidelines. We also discuss emerging developments like multimodal RAG (combining text with imaging or other data), ensemble model approaches, and new explainability tools that trace model outputs to sources. Finally, we critically analyze the limitations and challenges of deploying RAG in healthcare, including computational costs, retrieval errors, noise or conflicts in retrieved information, and ethical and regulatory considerations. While RAG-based systems show promise in augmenting oncologists’ expertise with timely knowledge, careful implementation, high-quality curation of knowledge bases, and human oversight will be crucial for safe and effective adoption in clinical practice.

Morel-Lavallee lesions (MLLs) are traumatic closed degloving injuries with operative intervention (OI), especially excisional debridement, remaining the mainstay of treatment. However, high-quality evidence describing operative techniques and outcomes remains limited. This study aimed to characterize current OI practices and associated outcomes, especially among patients undergoing excisional debridement, hypothesizing an increased risk of wound complications with this approach for MLLs. Methods This prospective (2021-2024), multicenter, observational study was conducted across 15 trauma centers and included patients diagnosed with MLL who underwent OI. The association between operative technique (eg, excisional debridement) and the primary outcome of wound complications was evaluated. Wound complications were defined as cellulitis, deep wound infection, skin necrosis, necrotizing soft tissue infection (NSTI), and/or organized hematoma. Results Of 88 MLL patients undergoing OI, 36 (40.9%) had excisional debridement, with 4 (11.1%) limited excision to skin, 20 (55.6%) extended to subcutaneous tissue, and 12 (33.3%) down to muscle. Excisional debridement had higher rates of skin necrosis (22.2% vs. 1.9%, P = .002). Although overall rates of wound complications were statistically similar between cohorts (33.3% vs. 17.3%, P = .098), univariate analysis demonstrated excisional debridement had increased risk of wound complications (OR 2.52, CI 1.07-5.97, P = .035). Conclusion This prospective multicenter study over three years demonstrated that a fifth of MLLs undergoing excisional debridement developed postoperative skin necrosis. Excisional debridement also had a more than two-fold increased risk of wound complications compared to other OIs. These findings underscore the importance of investigating alternative operative techniques, such as incision and drainage or minimally invasive debridement methods.

Background While rural health guidelines have often cautioned against digital interventions due to infrastructure and literacy barriers, this study introduces an SMS-based conversational artificial intelligence (CAI) that overcomes these limitations to promote HPV vaccination. Objective To evaluate an SMS-based CAI using the Centers for Disease Control (CDC)’s presumptive and CASE (Corroborate, About Me, Science and Explain) frameworks to promote HPV vaccination among rural parent-adolescent dyads by: (1) testing its usability; (2) assessing the feasibility of co-design with clinicians and parent-adolescents; (3) identifying adoption barriers and (4) examining communication preferences across behavioural profiles. Methods We developed a CAI simulator using the presumptive approach and the structured CASE method to address 11 common concerns about the HPV vaccine. A multimethod usability study with 13 rural parent-adolescent dyads (n=26) included in-depth interviews and the System Usability Scale. Results The CAI achieved excellent usability (mean SUS 85.2, ~96th percentile). A previously undocumented behavioural profile—vaccine-unaware—was identified, distinct from hesitancy, and best addressed through empathetic, exploratory dialogue. Key adoption barriers included integration with existing preventive care routines and endorsement by trusted local providers. Conclusions Contrary to the assumption that rural residents resist digital tools, participants engaged readily when content was credible, locally endorsed and accessible via SMS. Findings highlight the conditional effectiveness of the CDC’s presumptive style in CAIs—well-suited for vaccine-compliant users but less effective for sceptical or unaware groups, who, we argue, would respond better to conversational, empathetic tones.

CD25 is a subunit of the interleukin-2 (IL-2) receptor on T cells and natural killer (NK) cells. Acute leukemias with oncogenic tyrosine kinases often include CD25 + leukemia subpopulations, which portend poor clinical outcomes for patients; however, acute leukemia cells do not respond to IL-2. Here, we identified CD25 and its phosphorylation by protein kinase Cδ (PKCδ) as central elements of a feedback loop that stabilized fluctuations in oncogenic tyrosine kinase signaling in acute lymphoblastic and myeloid leukemia. Genetic ablation of CD25 in murine and patient-derived xenograft (PDX) models of acute leukemias reduced clonal fitness, colony formation, and leukemia-initiation capacity in serial transplant recipients. Oncogenic tyrosine kinase signaling in leukemia cells stimulated NF-κB–mediated CD25 expression, whereas PKCδ-mediated phosphorylation of CD25 suppressed oncogenic tyrosine kinase signaling through inhibitory phosphatases, such as PTPN6. Interactome analyses and mass spectrometry–based global phosphoproteomic analyses showed that CD25 deletion abolished the phosphatase activity of PTPN6, resulting in enhanced activation of tyrosine kinases and NF-κB. Four injections of a CD25 antibody-drug conjugate induced complete remission in mice transplanted with PDX refractory leukemia. These findings highlight the dependency of tyrosine kinase–driven leukemias on robust feedback control and the role of PKCδ and CD25 in assembling its components.

Abstract Cerebrospinal fluid (CSF) proteomics offers insights into molecular changes in aging and Alzheimer’s disease (AD). Key AD biomarkers, in particular amyloid-β (Aβ) and tau, in CSF are strongly associated with APOE genotype, the strongest genetic risk determinant of AD. To investigate how APOE genotype influences CSF proteome across AD pathology and age, we analyzed 362 neurology-related proteins and established AD biomarkers in CSF from 145 cognitively unimpaired participants in the Mayo Clinic Study of Aging. Importantly, our cohort is uniquely balanced across APOE genotypes, with similar representation of APOE2 carriers, APOE3/3 genotype, and APOE4 carriers. We identified several proteins, including lipid metabolism-related Lp-PLA2 and immune-related ITGAM, with strong APOE genotype-specific association. Notably, meta-analysis confirmed that ITGAM levels were consistently higher in APOE4 compared to APOE2 carriers across multiple cohorts and proteomic platforms. In addition, with increasing amyloid deposition, APOE4 carriers exhibited stronger immune responses, reflected by elevated ITGAM, TNF-α receptors, and IL-6, whereas APOE2 carriers showed attenuated responses. We further observed sex-specific effects among APOE2 carriers, characterized by distinct patterns in amyloid and CXCL11 levels. These findings suggest distinct mechanisms underlying APOE2’s protective and APOE4’s detrimental effects in brain aging and AD, paving for personalized diagnostics and interventions.

Abstract INTRODUCTION The genetic basis of sporadic early‐onset Alzheimer's disease (EOAD) remains largely unknown, prompting evaluation of late‐onset Alzheimer's disease (LOAD) polygenic risk in EOAD. METHODS A LOAD polygenic score (PGS) was calculated in the Longitudinal Early‐onset Alzheimer's Disease Study (LEADS) and Alzheimer's Disease Neuroimaging Initiative (ADNI) study and tested for associations with AD risk, cognitive performance, and imaging and fluid biomarkers. RESULTS Though PGS was elevated in LOAD and EOAD, it was not a significant predictor of EOAD adjusting for APOE ε4 carrier status and was not associated with age of EOAD onset ( p = 0.106) or with cognitive performance ( p = 0.417). In LEADS, greater LOAD PGS was associated with differences in neuroimaging and fluid biomarkers, including elevated synaptosomal‐associated protein 25 (SNAP‐25) ( p = 2.3 × 10 −5 ). DISCUSSION While LOAD polygenic risk contributed minimally to EOAD onset and cognitive dysfunction, PGS association with fluid biomarkers in LEADS suggests a role for LOAD polygenic risk in EOAD pathophysiology. Highlights LOAD PGSs were elevated in both LOAD and EOAD compared to controls; however, LOAD PGS did not significantly predict EOAD risk, age at onset, or cognitive performance independent of APOE ε4 in the LEADS. Higher LOAD PGS was associated with lower amyloid PET Centiloids (less brain amyloid deposition) as well as lower CSF biomarker Aβ42 in LEADS (proxy marker suggesting higher brain amyloid deposition) in LEADS; these contradictory findings support the need for larger studies to further investigate whether LOAD PGS is associated with increased amyloid deposition in EOAD. Higher LOAD PGS was also associated with higher levels of CSF synaptosomal‐associated protein 25 (SNAP‐25), a key component of the SNARE complex, suggesting that LOAD genetic factors may contribute to dysregulation of synaptic transmission and/or pathological protein aggregation in EOAD.

Abstract Secondary central nervous system (CNS) large B-cell lymphoma (SCNSL) occurs in the de novo setting, as a CNS-isolated relapse, or synchronous (concomitant CNS and systemic) relapse. SCNSL is a devastating event without therapeutic consensus. Thus, we aimed to evaluate treatment outcomes in an international cohort. Progression-free survival (PFS), overall survival (OS), and cumulative incidence of relapse (CIR, estimated using competing-risk models) were reported. Prognostic factors were identified in a 6-month landmark multivariate analysis. Outcomes after thiotepa autologous stem cell transplant (ASCT) and chimeric antigen receptor (CAR) T-cell therapy (CAR-T) delivered at relapse were compared after propensity score matching (PSM). A total of 1139 patients were included in the analysis (de novo: 537; relapsed SCNSL: 602). Two-year PFS estimates were 40.4%, 43.9%, and 16.2% for de novo SCNSL, CNS-isolated relapse, and synchronous relapse, respectively. Patients with CNS-isolated relapse demonstrated low rates of systemic recurrence (24-month CIR, 6%). Thiotepa-ASCT correlated with longer survival in de novo SCNSL (PFS: hazard ratio [HR], 0.57; P = .005; and OS: HR, 0.62; P = .023) and CNS-isolated relapses (PFS: HR, 0.55; P = .002; and OS: HR, 0.39; P< .0001). ASCT (thiotepa or no thiotepa) also associated with improved survival in synchronous relapses (PFS: HR, 0.57; P = .023; and OS: HR, 0.48; P = .019). Higher survival with thiotepa-ASCT than CAR-T was observed after PSM (PFS: HR, 0.45; P = .005 and OS: HR, 0.41; P = .014). These data support thiotepa-ASCT in eligible patients, particularly de novo disease and CNS-isolated relapses. CNS-isolated relapse was infrequently associated with systemic recurrence, supporting treatment regimens adopted from primary CNS lymphoma.