Search for: All records

Idiopathic pulmonary fibrosis (IPF) is a progressive, irreversible, and usually fatal lung disease of unknown reasons, generally affecting the elderly population. Early diagnosis of IPF is crucial for triaging patients’ treatment planning into anti‐fibrotic treatment or treatments for other causes of pulmonary fibrosis. However, current IPF diagnosis workflow is complicated and time‐consuming, which involves collaborative efforts from radiologists, pathologists, and clinicians and it is largely subject to inter‐observer variability.

The purpose of this work is to develop a deep learning‐based automated system that can diagnose subjects with IPF among subjects with interstitial lung disease (ILD) using an axial chest computed tomography (CT) scan. This work can potentially enable timely diagnosis decisions and reduce inter‐observer variability.

Our dataset contains CT scans from 349 IPF patients and 529 non‐IPF ILD patients. We used 80% of the dataset for training and validation purposes and 20% as the holdout test set. We proposed a two‐stage model: at stage one, we built a multi‐scale, domain knowledge‐guided attention model (MSGA) that encouraged the model to focus on specific areas of interest to enhance model explainability, including both high‐ and medium‐resolution attentions; at stage two, we collected the output from MSGA and constructed a random forest (RF) classifier for patient‐level diagnosis, to further boost model accuracy. RF classifier is utilized as a final decision stage since it is interpretable, computationally fast, and can handle correlated variables. Model utility was examined by (1) accuracy, represented by the area under the receiver operating characteristic curve (AUC) with standard deviation (SD), and (2) explainability, illustrated by the visual examination of the estimated attention maps which showed the important areas for model diagnostics.

During the training and validation stage, we observe that when we provide no guidance from domain knowledge, the IPF diagnosis model reaches acceptable performance (AUC±SD = 0.93±0.07), but lacks explainability; when including only guided high‐ or medium‐resolution attention, the learned attention maps are not satisfactory; when including both high‐ and medium‐resolution attention, under certain hyperparameter settings, the model reaches the highest AUC among all experiments (AUC±SD = 0.99±0.01) and the estimated attention maps concentrate on the regions of interests for this task. Three best‐performing hyperparameter selections according to MSGA were applied to the holdout test set and reached comparable model performance to that of the validation set.

Our results suggest that, for a task with only scan‐level labels available, MSGA+RF can utilize the population‐level domain knowledge to guide the training of the network, which increases both model accuracy and explainability.

 

Nan Laird has an enormous and growing impact on computational statistics. Her paper with Dempster and Rubin on the expectation‐maximisation (EM) algorithm is the second most cited paper in statistics. Her papers and book on longitudinal modelling are nearly as impressive. In this brief survey, we revisit the derivation of some of her most useful algorithms from the perspective of the minorisation‐maximisation (MM) principle. The MM principle generalises the EM principle and frees it from the shackles of missing data and conditional expectations. Instead, the focus shifts to the construction of surrogate functions via standard mathematical inequalities. The MM principle can deliver a classical EM algorithm with less fuss or an entirely new algorithm with a faster rate of convergence. In any case, the MM principle enriches our understanding of the EM principle and suggests new algorithms of considerable potential in high‐dimensional settings where standard algorithms such as Newton's method and Fisher scoring falter.

 

Linear mixed models are widely used for analyzing longitudinal datasets, and the inference for variance component parameters relies on the bootstrap method. However, health systems and technology companies routinely generate massive longitudinal datasets that make the traditional bootstrap method infeasible. To solve this problem, we extend the highly scalable bag of little bootstraps method for independent data to longitudinal data and develop a highly efficient Julia packageMixedModelsBLB.jl.Simulation experiments and real data analysis demonstrate the favorable statistical performance and computational advantages of our method compared to the traditional bootstrap method. For the statistical inference of variance components, it achieves 200 times speedup on the scale of 1 million subjects (20 million total observations), and is the only currently available tool that can handle more than 10 million subjects (200 million total observations) using desktop computers.

 

Tumor subtype and menopausal status are strong predictors of breast cancer (BC) prognosis. We aimed to find and validate subtype- or menopausal-status-specific changes in tumor DNA methylation (DNAm) associated with all-cause mortality or BC progression. Associations between site-specific tumor DNAm and BC prognosis were estimated among The Cancer Genome Atlas participants ( n = 692) with Illumina Infinium HumanMethylation450 BeadChip array data. All-cause mortality and BC progression were modeled using Cox proportional hazards models stratified by tumor subtypes, adjusting for age, race, stage, menopausal status, tumor purity, and cell type proportion. Effect measure modification by subtype and menopausal status were evaluated by incorporating a product term with DNAm. Site-specific inference was used to identify subtype- or menopausal-status-specific differentially methylated regions (DMRs) and functional pathways. The validation of the results was carried out on an independent dataset (GSE72308; n = 180). We identified a total of fifteen unique CpG probes that were significantly associated ( P ≤ 1 × 10 − 7 with survival outcomes in subtype- or menopausal-status-specific manner. Seven probes were associated with overall survival (OS) or progression-free interval (PFI) for women with luminal A subtype, and four probes were associated with PFI for women with luminal B subtype. Five probes were associated with PFI for post-menopausal women. A majority of significant probes showed a lower risk of OS or BC progression with higher DNAm. We identified subtype- or menopausal-status-specific DMRs and functional pathways of which top associated pathways differed across subtypes or menopausal status. None of significant probes from site-specific analyses met genome-wide significant level in validation analyses while directions and magnitudes of coefficients showed consistent pattern. We have identified subtype- or menopausal-status-specific DNAm biomarkers, DMRs and functional pathways associated with all-cause mortality or BC progression, albeit with limited validation. Future studies with larger independent cohort of non-post-menopausal women with non-luminal A subtypes are warranted for identifying subtype- and menopausal-status-specific DNAm biomarkers for BC prognosis. 

Abstract Objective Modern healthcare data reflect massive multi-level and multi-scale information collected over many years. The majority of the existing phenotyping algorithms use case–control definitions of disease. This paper aims to study the time to disease onset and progression and identify the time-varying risk factors that drive them. Materials and Methods We developed an algorithmic approach to phenotyping the incidence of diseases by consolidating data sources from the UK Biobank (UKB), including primary care electronic health records (EHRs). We focused on defining events, event dates, and their censoring time, including relevant terms and existing phenotypes, excluding generic, rare, or semantically distant terms, forward-mapping terminology terms, and expert review. We applied our approach to phenotyping diabetes complications, including a composite cardiovascular disease (CVD) outcome, diabetic kidney disease (DKD), and diabetic retinopathy (DR), in the UKB study. Results We identified 49 049 participants with diabetes. Among them, 1023 had type 1 diabetes (T1D), and 40 193 had type 2 diabetes (T2D). A total of 23 833 diabetes subjects had linked primary care records. There were 3237, 3113, and 4922 patients with CVD, DKD, and DR events, respectively. The risk prediction performance for each outcome was assessed, and our results are consistent with the prediction area under the ROC (receiver operating characteristic) curve (AUC) of standard risk prediction models using cohort studies. Discussion and Conclusion Our publicly available pipeline and platform enable streamlined curation of incidence events, identification of time-varying risk factors underlying disease progression, and the definition of a relevant cohort for time-to-event analyses. These important steps need to be considered simultaneously to study disease progression. 

Introduction Studies have reported that antidiabetic medications (ADMs) were associated with lower risk of dementia, but current findings are inconsistent. This study compared the risk of dementia onset in patients with type 2 diabetes (T2D) treated with sulfonylurea (SU) or thiazolidinedione (TZD) to patients with T2D treated with metformin (MET). Research design and methods This is a prospective observational study within a T2D population using electronic medical records from all sites of the Veterans Affairs Healthcare System. Patients with T2D who initiated ADM from January 1, 2001, to December 31, 2017, were aged ≥60 years at the initiation, and were dementia-free were identified. A SU monotherapy group, a TZD monotherapy group, and a control group (MET monotherapy) were assembled based on prescription records. Participants were required to take the assigned treatment for at least 1 year. The primary outcome was all-cause dementia, and the two secondary outcomes were Alzheimer’s disease and vascular dementia, defined by International Classification of Diseases (ICD), 9th Revision, or ICD, 10th Revision, codes. The risks of developing outcomes were compared using propensity score weighted Cox proportional hazard models. Results Among 559 106 eligible veterans (mean age 65.7 (SD 8.7) years), the all-cause dementia rate was 8.2 cases per 1000 person-years (95% CI 6.0 to 13.7). After at least 1 year of treatment, TZD monotherapy was associated with a 22% lower risk of all-cause dementia onset (HR 0.78, 95% CI 0.75 to 0.81), compared with MET monotherapy, and 11% lower for MET and TZD dual therapy (HR 0.89, 95% CI 0.86 to 0.93), whereas the risk was 12% higher for SU monotherapy (HR 1.12 95% CI 1.09 to 1.15). Conclusions Among patients with T2D, TZD use was associated with a lower risk of dementia, and SU use was associated with a higher risk compared with MET use. Supplementing SU with either MET or TZD may partially offset its prodementia effects. These findings may help inform medication selection for elderly patients with T2D at high risk of dementia.