Laila M. Poisson

Computational competitions are the standard for benchmarking medical image analysis algorithms, but they typically use small curated test datasets acquired at a few centers, leaving a gap to the reality of diverse multicentric patient data. To this end, the Federated Tumor Segmentation (FeTS) Challenge represents the paradigm for real-world algorithmic performance evaluation. The FeTS challenge is a competition to benchmark (i) federated learning aggregation algorithms and (ii) state-of-the-art segmentation algorithms, across multiple international sites. Weight aggregation and client selection techniques were compared using a multicentric brain tumor dataset in realistic federated learning simulations, yielding benefits for adaptive weight aggregation, and efficiency gains through client sampling. Quantitative performance evaluation of state-of-the-art segmentation algorithms on data distributed internationally across 32 institutions yielded good generalization on average, albeit the worst-case performance revealed data-specific modes of failure. Similar multi-site setups can help validate the real-world utility of healthcare AI algorithms in the future.

PURPOSE: Patient-reported outcome measures (PROMs) provide a direct report of the patient's perspective, complementary to clinician assessment. Currently, understanding the real-time changes in PROM scores near the end of life remains limited. This study evaluated differences in mean PROM scores between patients with cancer within 6 months before death compared with surviving patients with cancer. METHODS: This retrospective case-control study uses the National Institutes of Health's Patient-Reported Outcomes Measurement Information System computer adaptive testing instruments to assess pain interference, physical function, fatigue, and depression. Patients dying within 6 months of PROM completion were selected as cases and matched to controls 1:3 by age at PROM completion, sex, cancer disease site, and cancer stage at diagnosis. Generalized estimating equation models assessed the difference in mean PROM score in cases compared with controls. RESULTS: A total of 461 cases and 1,270 controls from September 2020 to January 2023 were included. After adjustment for ethnicity, Charlson Comorbidity Index, and census tract median household income, significant differences in mean scores were demonstrated. Physical function domain showed the largest difference, with cases averaging 6.52 points lower than controls (95% CI, –8.25 to –4.80). Fatigue and pain interference domains showed a rise in PROMs scores by 4.83 points (95% CI, 2.94 to 6.72) and 4.33 points (95% CI, 2.53 to 6.12), respectively. CONCLUSION: Compared with controls, patients dying within 6 months of PROM completion demonstrated worse PROM scores in the four domains assessed. These findings suggest the utility of routinely collected PROMs as a real-time indicator of the terminal stage of life among patients with cancer to allow for earlier intervention with supportive oncology services.

Glioblastoma (GBM) tumors exhibit extensive genomic, epigenomic, and transcriptional diversity, with significant intratumoral heterogeneity, complicating standard treatment approaches involving radiation (RT) and the DNA-alkylating agent temozolomide (TMZ). In this study, we employed an integrative multi-omics approach, including targeted proteomics, transcriptomics, genomics, and DNA methylation profiling, to investigate the response of a representative panel of GBM patient-derived cancer stem cells (CSCs) to astrocytic differentiation and RT and TMZ treatments. Differentiated CSC progenies retained the expression of key stemness genes and survival pathways, while activating the BMP-Smad signaling pathway and upregulating extracellular matrix components. This was associated with increased resistance to TMZ, though not to RT, across all models. We identified TP53 status as a critical determinant of transcriptional response to both RT and TMZ, which was also modulated by the differentiation state and treatment modality in wildtype (wt) p53 GBM cells. Both mutant and wt p53 models exhibited significant activation of the DNA-damage associated interferon (IFN) response in CSCs and differentiated cells, implicating this pathway in the GBM response to therapy. We observed that activation of NF-κB was positively correlated with the levels of O-6-methylguanine-DNA methyltransferase (MGMT) protein, a direct DNA repair enzyme leading to TMZ resistance, regardless of MGMT promoter methylation status, further supporting the clinical potential for inhibition of NF-kB signaling in GBM treatment. Our integrative analysis of the impact of GBM cell developmental states, in the context of genomic and molecular diversity of patient-derived models, provides valuable insights for pre-clinical studies aimed at optimizing treatment strategies.

Background: Chronic insomnia disorder affects 10–15% of adults, causing significant individual and societal burden. Despite Cognitive Behavioral Therapy for Insomnia (CBT-I) being the recommended first-line, sleep medications remain more common due to limited access to trained providers. Digital CBT-I offers a scalable solution, but evidence of its real-world impact in U.S. clinical settings is lacking. Method: This study evaluates real-world implementation and impact of digital CBT-I in U.S. clinical settings, using Normalization Process Theory (NPT) to guide integration at Henry Ford Health, Detroit, Michigan. Implementation success was assessed through order rates, patient sign-ups and workflow acceptability. We assess the effect on healthcare utilization through a propensity-matched observational treatment-control design. Results: Implementation was successful, with 1,162 patients offered digital CBT-I. From this cohort, we analyzed a sample of 340 patients with sufficient chart data and established care (120 days) who utilized digital CBT-I, comparing them to 340 matched standard care controls. Patients who used digital CBT-I had a 64% reduction in the odds of any medication fill during the postwindow period ( p < .001) and were 53% less likely to fill insomnia medication prescriptions compared with the preperiod ( p = .013). Controls did not have any significant reductions in medication fill rates. Time-varied analysis showed digital CBT-I patients had transiently higher outpatient visit odds at 30–60 days, followed by sustained reductions of 28% (120–150 days) and 31% (150–180 days). After covariate adjustment, early differences were nonsignificant while later reductions remained significant. Conclusions: NPT facilitated integration of digital CBT-I into existing workflows, allowing immediate access while minimizing disruption to routine practice. Provider training sessions and reminders effectively promoted suitable patient uptake. Digital CBT-I was associated with reduced medication fills pre-to-post with an initial rise and then sustained reduction in outpatient service utilization patterns over time. A key limitation is the use of individuals who declined digital CBT-I as comparators, which may introduce selection bias. Generalizability may be limited as the study was conducted within a single healthcare system. Trial Registration: Not applicable—the assignment of the medical intervention to patients was not at the discretion of the investigators.

Multiple sclerosis (MS) is the most common inflammatory neurodegenerative disease of the central nervous system (CNS) in young adults and results in progressive neurological defects. The relapsing-remitting phenotype (RRMS) is the most common disease course in MS, which ultimately progresses to secondary progressive MS (SPMS), while primary progressive MS (PPMS) is a type of MS that worsens gradually over time without remissions. There is a gap in knowledge regarding whether the relapsing form can be distinguished from the progressive course, or healthy subjects (HS) based on an altered serum metabolite profile. In this study, we performed global untargeted metabolomics with the 2D GC-GC-MS platform to identify altered metabolites between RRMS, PPMS, and HS. We profiled 235 metabolites in the serum of patients with RRMS (n = 41), PPMS (n = 31), and HS (n = 91). A comparison of RRMS and HS patients revealed 22 significantly altered metabolites at p < 0.05 (false-discovery rate [FDR] = 0.3). The PPMS and HS comparisons revealed 28 altered metabolites at p < 0.05 (FDR = 0.2). Pathway analysis using MetaboAnalyst revealed enrichment of four metabolic pathways in both RRMS and PPMS (hypergeometric test p < 0.05): (1) galactose metabolism; (2) amino sugar and nucleotide sugar metabolism; (3) phenylalanine, tyrosine, and tryptophan biosynthesis; and (4) aminoacyl-tRNA biosynthesis. The Qiagen IPA enrichment test identified the sulfatase 2 (SULF2) (p = 0.0033) and integrin subunit beta 1 binding protein 1 (ITGB1BP1) (p = 0.0067) genes as upstream regulators of altered metabolites in the RRMS vs. HS groups. However, in the PPMS vs. HS comparison, valine was enriched in the neurodegeneration of brain cells (p = 0.05), and heptadecanoic acid, alpha-ketoisocaproic acid, and glycerol participated in inflammation in the CNS (p = 0.03). Overall, our study suggests that RRMS and PPMS may contribute metabolic fingerprints in the form of unique altered metabolites for discriminating MS disease from HS, with the potential for constructing a metabolite panel for progressive autoimmune diseases such as MS.

While OpenAI’s GPT-4 with vision demonstrated a level of competency in answering text-based radiology in-training examination questions, the model showed deficits in analyzing key radiologic images.

Data‐intensive research continues to expand with the goal of improving healthcare delivery, clinical decision‐making, and patient outcomes. Quantitative scientists, such as biostatisticians, epidemiologists, and informaticists, are tasked with turning data into health knowledge. In academic health centres, quantitative scientists are critical to the missions of biomedical discovery and improvement of health. Many academic health centres have developed centralized Quantitative Science Units which foster dual goals of professional development of quantitative scientists and producing high quality, reproducible domain research. Such units then develop teams of quantitative scientists who can collaborate with researchers. However, existing literature does not provide guidance on how such teams are formed or how to manage and sustain them. Leaders of Quantitative Science Units across six institutions formed a working group to examine common practices and tools that can serve as best practices for Quantitative Science Units that wish to achieve these dual goals through building long‐term partnerships with researchers. The results of this working group are presented to provide tools and guidance for Quantitative Science Units challenged with developing, managing, and evaluating Quantitative Science Teams. This guidance aims to help Quantitative Science Units effectively participate in and enhance the research that is conducted throughout the academic health centre—shaping their resources to fit evolving research needs.

Background: Understanding sex-based differences in glioblastoma patients is necessary for accurate personalized treatment planning to improve patient outcomes. Purpose: To investigate sex-specific differences in molecular, clinical and radiological tumor parameters, as well as survival outcomes in glioblastoma, isocitrate dehydrogenase-1 wildtype (IDH1-WT), grade 4 patients. Methods: Retrospective data of 1832 glioblastoma, IDH1-WT patients with comprehensive information on tumor parameters was acquired from the Radiomics Signatures for Precision Oncology in Glioblastoma (ReSPOND) consortium. Data imputation was performed for missing values. Sex-based differences in tumor parameters, such as, age, molecular parameters, pre-operative KPS score, tumor volumes, epicenter and laterality were assessed through non-parametric tests. Spatial atlases were generated using pre-operative MRI maps to visualize tumor characteristics. Survival time analysis was performed through log-rank tests and Cox proportional hazard analyses. Results: GBM was diagnosed at a median age of 64 years in females compared to 61.9 years in males (FDR = 0.003). Males had a higher Karnofsky Performance Score (above 80) as compared to females (60.4% females Vs 69.7% males, FDR = 0.044). Females had lower tumor volumes in enhancing (16.7 cm³ Vs. 20.6 cm³ in males, FDR = 0.001), necrotic core (6.18 cm³ Vs. 7.76 cm³ in males, FDR = 0.001) and edema regions (46.9 cm³ Vs. 59.2 cm³ in males, FDR = 0.0001). Right temporal region was the most common tumor epicenter in the overall population. Right as well as left temporal lobes were more frequently involved in males. There were no significant differences in survival outcomes and mortality ratios. Higher age, unmethylated O6-methylguanine-DNAmethyltransferase (MGMT) promoter and undergoing subtotal resection increased the mortality risk in both males and females. Conclusions: Our study demonstrates significant sex-based differences in clinical and radiological tumor parameters of glioblastoma, IDH1-WT, grade 4 patients. Sex is not an independent prognostic factor for survival outcomes and the tumor parameters influencing patient outcomes are identical for males and females. ABBREVIATIONS: IDH1-WT = isocitrate dehydrogenase-1 wildtype; MGMTp = O6-methylguanine-DNA-methyltransferase promoter; KPS = Karnofsky performance score; EOR = extent of resection; WHO = world health organization; FDR = false discovery rate

Glioblastoma (GBM) tumors represents diverse genomic epigenomic, and transcriptional landscapes, with significant intratumoral heterogeneity that challenges standard of care treatments involving radiation (RT) and the DNA-alkylating agent temozolomide (TMZ). In this study, we employed targeted proteomics to assess the response of a genomically-diverse panel of GBM patient-derived cancer stem cells (CSCs) to astrocytic differentiation, growth factor withdrawal and traditional high fetal bovine serum culture. Our findings revealed a complex crosstalk and co-activation of key oncogenic signaling in CSCs and diverse patterns of response to these external stimuli. Using RNA sequencing and DNA methylation, we observed common adaptations in response to astrocytic differentiation of CSCs across genomically distinct models, including BMP-Smad pathway activation, reduced cholesterol biosynthesis, and upregulation of extracellular matrix components. Notably, we observed that these differentiated CSC progenies retained a subset of stemness genes and the activation of cell survival pathways. We also examined the impact of differentiation state and genomic background on GBM cell sensitivity and transcriptional response to TMZ and RT. Differentiation of CSCs increased resistance to TMZ but not to RT. While transcriptional responses to these treatments were predominantly regulated by p53 in wild-type p53 GBM cells, its transcriptional activity was modulated by the differentiation status and treatment modality. Both mutant and wild-type p53 models exhibited significant activation of a DNA-damage associated interferon response in CSCs and differentiated cells, suggesting this pathway may play a wider role in GBM response to TMZ and RT. Our integrative analysis of the impact of GBM cell developmental states, in the context of genomic and molecular diversity of patient-derived models, provides valuable insights for pre-clinical studies aimed at optimizing treatment strategies.