The textures of sarcoidosis: quantifying lung disease through variograms.

Objective: Sarcoidosis is a granulomatous disease affecting the lungs in over 90% of patients. Qualitative assessment of chest CT by radiologists is standard clinical practice and reliable quantification of disease from CT would support ongoing efforts to identify sarcoidosis phenotypes. Standard imaging feature engineering techniques such as radiomics suffer from extreme sensitivity to image acquisition and processing, potentially impeding generalizability of research to clinical populations. In this work, we instead investigate approaches to engineering variogram-based features with the intent to identify a robust, generalizable pipeline for image quantification in the study of sarcoidosis. Approach: For a cohort of more than 300 individuals with sarcoidosis, we investigated 24 feature engineering pipelines differing by decisions for image registration to a template lung, empirical and model variogram estimation methods, and feature harmonization for CT scanner model, and subsequently 48 sets of phenotypes produced through unsupervised clustering. We then assessed sensitivity of engineered features, phenotypes produced through unsupervised clustering, and sarcoidosis disease signal strength to pipeline. Main results: We found that variogram features had low to mild association with scanner model and associations were reduced by image registration. For each feature type, features were also typically robust to all pipeline decisions except image registration. Strength of disease signal as measured by association with pulmonary function testing and some radiologist visual assessments was strong (optimistic AUC ≈ 0.9, p ≪ 0.0001 in models for architectural distortion, conglomerate mass, fibrotic abnormality, and traction bronchiectasis) and fairly consistent across engineering approaches regardless of registration and harmonization for CT scanner. Significance: Variogram-based features appear to be a suitable approach to image quantification in support of generalizable research in pulmonary sarcoidosis.

Chest computed tomography provides more information than chest X-ray alone in determining extent of physiologic impairment in pulmonary sarcoidosis.

BACKGROUND: Sarcoidosis staging has primarily relied on the Scadding chest radiographic system, although chest CT is finding increased clinical use. RESEARCH QUESTION: Whether standardized CT assessment provides additional understanding of lung function beyond Scadding stage and demographics is unknown and the focus of this study. STUDY DESIGN AND METHODS: We used the NHLBI study Genomics Research in Alpha-1 Anti-Trypsin Deficiency and Sarcoidosis (GRADS) sarcoidosis cases(N=351) with Scadding stage and Chest CT scans obtained in a standardized manner. One chest radiologist scored all CT scans with a visual scoring system, with a subset read by another chest radiologist. We compared demographic features, Scadding stage and CT findings, and correlation between these measures. Associations between spirometry and DLCO and CT and Scadding stage were determined using regression analysis (N=318). Agreement between readers was evaluated using Cohen's Kappa. RESULTS: CT features were inconsistent with Scadding stage in about ∼40% of cases. Most CT features assessed on visual scoring were negatively associated with lung function. Associations persisted for FEV1 and DLCO when adjusting for Scadding stage, although some CT feature associations with FVC became insignificant. Scadding stage was primarily associated with FEV1 and inclusion of CT features reduced significance in association between Scadding and lung function. Multivariable regression modeling to identify radiologic measures explaining lung function included Scadding stage for FEV1 and FEV1/FVC (P<0.05) and marginally for DLCO (P<0.15). Combinations of CT measures accounted for Scadding stage for FVC. Correlations among Scadding and CT features were noted. Agreement between readers was poor to moderate for presence/absence of CT features and poor for degree/location of abnormality. INTERPRETATION: CT features explained additional variability in lung function beyond Scadding stage, with some CT features obviating the associations between lung function and Scadding. Whether CT features/phenotypes/endotypes could be useful for managing patients with sarcoidosis needs more study.

Developmental drugs for sarcoidosis.

Sarcoidosis is a multi-organ granulomatous inflammatory disease of unknown etiology. Over 50% of patients will require treatment at some point in their disease and 10%-30% will develop a chronic progressive disease with pulmonary fibrosis leading to significant morbidity and mortality. Recently published guidelines recommend immunosuppressive therapy for sarcoidosis patients at risk of increased disease-related morbidity and mortality, and in whom disease has negatively impacted quality of life. Prednisone the currently recommended first line therapy is associated with significant toxicity however none of the other guideline recommended steroid sparing therapy is approved by regulatory agencies for use in sarcoidosis, and data in support of their use is weak. For patients with severe refractory disease requiring prolonged therapy, treatment options are limited. The need for expanding treatment options in sarcoidosis has been emphasized. Well conducted large, randomized trials evaluating currently available therapeutic options as well as novel pathways for targeting disease are necessary to better guide treatment decisions. These trials will not be without significant challenges. Sarcoidosis is a rare disease with heterogenous presentation and variable progression and clinical outcome. There are no universally agreed upon biomarkers of disease activity and measurement of outcomes is confounded by the need to balance patient centric measures and objective measures of disease activity. Our paper provides an update on developmental drugs in sarcoidosis and outlines several novel pathways that may be targeted for future drug development. Currently available trials are highlighted and ongoing challenges to drug development and clinical trial design are briefly discussed.

Compartment-specific protein interactions in beryllium lung disease.

The study provides insights into proteins that may be relevant in BeS and CBD. It provides a framework to investigate the global changes in lung compartment-specific inflammatory cells to better understand the potential interplay of proteins in CBD. https://bit.ly/3PLNTXC.

Genome-wide association study identifies multiple HLA loci for sarcoidosis susceptibility.

Sarcoidosis is a complex systemic disease. Our study aimed to (1) identify novel alleles associated with sarcoidosis susceptibility; (2) provide an in-depth evaluation of HLA alleles and sarcoidosis susceptibility and (3) integrate genetic and transcription data to identify risk loci that may more directly impact disease pathogenesis. We report a genome-wide association study of 1335 sarcoidosis cases and 1264 controls of European descent (EA) and investigate associated alleles in a study of African Americans (AA: 1487 cases and 1504 controls). The EA and AA cohort was recruited from multiple United States sites. HLA alleles were imputed and tested for association with sarcoidosis susceptibility. Expression quantitative locus and colocalization analysis were performed using a subset of subjects with transcriptome data. Forty-nine SNPs in the HLA region in HLA-DRA, -DRB9, -DRB5, -DQA1 and BRD2 genes were significantly associated with sarcoidosis susceptibility in EA, rs3129888 was also a risk variant for sarcoidosis in AA. Classical HLA alleles DRB1*0101, DQA1*0101 and DQB1*0501, which are highly correlated, were also associated with sarcoidosis. rs3135287 near HLA-DRA was associated with HLA-DRA expression in peripheral blood mononuclear cells and bronchoalveolar lavage from subjects and lung tissue and whole blood from GTEx. We identified six novel SNPs (out of the seven SNPs representing the 49 significant SNPs) and nine HLA alleles associated with sarcoidosis susceptibility in the largest EA population. We also replicated our findings in an AA population. Our study reiterates the potential role of antigen recognition and/or presentation HLA class II genes in sarcoidosis pathogenesis.

Standardization of flow cytometry and cell sorting to enable a transcriptomic analysis in a multi-site sarcoidosis study.

The contribution and regulation of various CD4+ T cell lineages that occur with remitting vs progressive courses in sarcoidosis are poorly understood. We developed a multiparameter flow cytometry panel to sort these CD4+ T cell lineages followed by measurement of their functional potential using RNA-sequencing analysis at six-month intervals across multiple study sites. To obtain good quality RNA for sequencing, we relied on chemokine receptor expression to identify and sort lineages. To minimize gene expression changes induced by perturbations of T cells and avoid protein denaturation caused by freeze/thaw cycles, we optimized our protocols using freshly isolated samples at each study site. To accomplish this study, we had to overcome significant standardization challenges across multiple sites. Here, we detail standardization considerations for cell processing, flow staining, data acquisition, sorting parameters, and RNA quality control analysis that were performed as part of the NIH-sponsored, multi-center study, BRonchoscopy at Initial sarcoidosis diagnosis Targeting longitudinal Endpoints (BRITE). After several rounds of iterative optimization, we identified the following aspects as critical for successful standardization: 1) alignment of PMT voltages across sites using CS&T/rainbow bead technology; 2) a single template created in the cytometer program that was used by all sites to gate cell populations during data acquisition and cell sorting; 3) use of standardized lyophilized flow cytometry staining cocktails to reduce technical error during processing; 4) development and implementation of a standardized Manual of Procedures. After standardization of cell sorting, we were able to determine the minimum number of sorted cells necessary for next generation sequencing through analysis of RNA quality and quantity from sorted T cell populations. Overall, we found that implementing a multi-parameter cell sorting with RNA-seq analysis clinical study across multiple study sites requires iteratively tested standardized procedures to ensure comparable and high-quality results.

Multi-Omic Signatures of Sarcoidosis and Progression in Bronchoalveolar Lavage Cells.

Introduction: Sarcoidosis is a heterogeneous, granulomatous disease that can prove difficult to diagnose, with no accurate biomarkers of disease progression. Therefore, we profiled and integrated the DNA methylome, mRNAs, and microRNAs to identify molecular changes associated with sarcoidosis and disease progression that might illuminate underlying mechanisms of disease and potential genomic biomarkers. Methods: Bronchoalveolar lavage cells from 64 sarcoidosis subjects and 16 healthy controls were used. DNA methylation was profiled on Illumina HumanMethylationEPIC arrays, mRNA by RNA-sequencing, and miRNAs by small RNA-sequencing. Linear models were fit to test for effect of diagnosis and phenotype, adjusting for age, sex, and smoking. We built a supervised multi-omics model using a subset of features from each dataset. Results: We identified 46,812 CpGs, 1,842 mRNAs, and 5 miRNAs associated with sarcoidosis versus controls and 1 mRNA, SEPP1 - a protein that supplies selenium to cells, associated with disease progression. Our integrated model emphasized the prominence of the PI3K/AKT1 pathway in sarcoidosis, which is important in T cell and mTOR function. Novel immune related genes and miRNAs including LYST, RGS14, SLFN12L, and hsa-miR-199b-5p, distinguished sarcoidosis from controls. Our integrated model also demonstrated differential expression/methylation of IL20RB, ABCC11, SFSWAP, AGBL4, miR-146a-3p, and miR-378b between non-progressive and progressive sarcoidosis. Conclusions: Leveraging the DNA methylome, transcriptome, and miRNA-sequencing in sarcoidosis BAL cells, we detected widespread molecular changes associated with disease, many which are involved in immune response. These molecules may serve as diagnostic/prognostic biomarkers and/or drug targets, although future testing will be required for confirmation.

Efzofitimod for the Treatment of Pulmonary Sarcoidosis.

BACKGROUND: Pulmonary sarcoidosis is characterized by the accumulation of immune cells that form granulomas affecting the lungs. Efzofitimod (ATYR1923), a novel immunomodulator, selectively binds neuropilin 2, which is upregulated on immune cells in response to lung inflammation. RESEARCH QUESTION: What is the tolerability, safety, and effect on outcomes of efzofitimod in pulmonary sarcoidosis? STUDY DESIGN AND METHODS: In this randomized, double-blind, placebo-controlled study evaluating multiple ascending doses of efzofitimod administered intravenously every 4 weeks for 24 weeks, randomized patients (2:1) underwent a steroid taper to 5 mg/d by week 8 or < 5 mg/d after week 16. The primary end point was the incidence of adverse events (AEs); secondary end points included steroid reduction, change in lung function, and patient-reported outcomes on health-related quality-of-life scales. RESULTS: Thirty-seven patients received at least one dose of study medication. Efzofitimod was well tolerated at all doses, with no new or unexpected AEs and no dose-dependent AE incidence. Average daily steroid doses through end of study were 6.8 mg, 6.5 mg, and 5.6 mg for the 1 mg/kg, 3 mg/kg, and 5 mg/kg groups compared with 7.2 mg for placebo, resulting in a baseline-adjusted relative steroid reduction of 5%, 9%, and 22%, respectively. Clinically meaningful improvements were achieved across several patient-reported outcomes, several of which reached statistical significance in the 5 mg/kg dose arm. A dose-dependent but nonsignificant trend toward improved lung function also was observed for 3 and 5 mg/kg. INTERPRETATION: Efzofitimod was safe and well tolerated and was associated with dose-dependent improvements of several clinically relevant end points compared with placebo. The results of this study support further evaluation of efzofitimod in pulmonary sarcoidosis. TRIAL REGISTRY: ClinicalTrials.gov; No.: NCT03824392; URL: www. CLINICALTRIALS: gov.

Examination of eQTL Polymorphisms Associated with Increased Risk of Progressive Complicated Sarcoidosis in European and African Descent Subjects.

Background: A limited pool of SNPs are linked to the development and severity of sarcoidosis, a systemic granulomatous inflammatory disease. By integrating genome-wide association studies (GWAS) data and expression quantitative trait loci (eQTL) single nuclear polymorphisms (SNPs), we aimed to identify novel sarcoidosis SNPs potentially influencing the development of complicated sarcoidosis. Methods: A GWAS (Affymetrix 6.0) involving 209 African-American (AA) and 193 European-American (EA, 75 and 51 complicated cases respectively) and publicly-available GWAS controls (GAIN) was utilized. Annotation of multi-tissue eQTL SNPs present on the GWAS created a pool of ~46,000 eQTL SNPs examined for association with sarcoidosis risk and severity (Logistic Model, Plink). The most significant EA/AA eQTL SNPs were genotyped in a sarcoidosis validation cohort (n=1034) and cross-validated in two independent GWAS cohorts. Results: No single GWAS SNP achieved significance (p<1x10-8), however, analysis of the eQTL/GWAS SNP pool yielded 621 eQTL SNPs (p<10-4) associated with 730 genes that highlighted innate immunity, MHC Class II, and allograft rejection pathways with multiple SNPs validated in an independent sarcoidosis cohort (105 SNPs analyzed) (NOTCH4, IL27RA, BTNL2, ANXA11, HLA-DRB1). These studies confirm significant association of eQTL/GWAS SNPs in EAs and AAs with sarcoidosis risk and severity (complicated sarcoidosis) involving HLA region and innate immunity. Conclusion: Despite the challenge of deciphering the genetic basis for sarcoidosis risk/severity, these results suggest that integrated eQTL/GWAS approaches may identify novel variants/genes and support the contribution of dysregulated innate immune responses to sarcoidosis severity.

Autoantibody profile in sarcoidosis, analysis from the GRADS sarcoidosis cohort.

BACKGROUND: Sarcoidosis, a multi-systemic granulomatous disease, is a predominantly T-cell disease but evidence for a role for humoral immunity in disease pathogenesis is growing. Utilizing samples from the Genomic Research in Alpha-1 anti-trypsin Deficiency and Sarcoidosis (GRADS) study, we examined the prevalence of autoantibodies in sarcoidosis patients with pulmonary-only and extra-pulmonary organ involvement compared to normal controls. STUDY DESIGN AND METHODS: We analyzed serum samples from sarcoidosis patients who participated in the GRADS study utilizing an autoantigen microarray platform for both IgM and IgG antibodies. The cohort included sarcoidosis patients with pulmonary-only disease (POS, n = 106), sarcoidosis patients with extra-pulmonary disease (EPS, n = 120) and a normal control cohort (NC, n = 101). Organ involvement was assessed following a standardized format across all GRADS participating centers. RESULTS: Sarcoidosis patients overall had increased levels of IgM and IgG autoantibodies compared to normal controls. In addition, several autoantibodies were elevated in the POS and EPS cohorts compared to the NC cohort. Differences in autoantibody levels were also noted between the POS and the EPS cohorts. When comparing organ involvement with sarcoidosis, bone, spleen and ear, nose and throat involvement had higher IgM expression than other organs. CONCLUSION: Sarcoidosis patients have elevated IgM and IgG autoantibody levels compared to normal controls. In addition, individuals with pulmonary as well as additional organ involvement had higher IgM expression. Further research is needed focusing on specific organ-autoantibody pairs and role of autoantibodies in disease pathogenesis.

Occupational and environmental exposures in the Genomic Research in Alpha-1 Antitrypsin Deficiency and Sarcoidosis (GRADS) study.

INTRODUCTION: Sarcoidosis is a granulomatous disorder thought to be caused by exposures in genetically susceptible individuals. This study investigated whether specific exposures were associated with different sarcoidosis phenotypes. METHODS: Extensive demographic, occupational and environmental exposure data was analyzed from subjects enrolled in the NHLBI Genomic Research in Alpha-1 Antitrypsin Deficiency and Sarcoidosis (GRADS) study. RESULTS: In patients with sarcoidosis, radiation exposure was significantly associated with an increased risk of cardiac sarcoidosis versus non-cardiac sarcoidosis. No exposures were significantly associated with pulmonary only disease versus extrapulmonary disease with or without pulmonary involvement, Scadding Stage II/III/IV versus Scadding Stage 0/I, acute or remitting disease versus non-acute or non-remitting disease, nor chronic versus non-chronic disease. Although not reaching statistically significance after adjustment for multiple comparisons, there were a number of exposures associated with specific disease phenotypes, including exposures where relationships to sarcoidosis have previously been described such as rural exposures and pesticide exposures. CONCLUSIONS: Radiation exposure may be a risk factor for cardiac sarcoidosis. Other exposures may also be associated with specific phenotypes and should be further explored. The study was limited by small groups of exposed subjects for individual exposures and multiple comparisons. The development of novel and innovative exposure assessment tools is needed.

Multiomic Signatures of Chronic Beryllium Disease Bronchoalveolar Lavage Cells Relate to T-Cell Function and Innate Immunity.

Chronic beryllium disease (CBD) is a Th1 granulomatous lung disease preceded by sensitization to beryllium (BeS). We profiled the methylome, transcriptome, and selected proteins in the lung to identify molecular signatures and networks associated with BeS and CBD. BAL cell DNA and RNA were profiled using microarrays from CBD (n = 30), BeS (n = 30), and control subjects (n = 12). BAL fluid proteins were measured using Olink Immune Response Panel proteins from CBD (n = 22) and BeS (n = 22) subjects. Linear models identified features associated with CBD, adjusting for covariation and batch effects. Multiomic integration methods identified correlated features between datasets. We identified 1,546 differentially expressed genes in CBD versus control subjects and 204 in CBD versus BeS. Of the 101 shared transcripts, 24 have significant cis relationships between gene expression and DNA methylation, assessed using expression quantitative trait methylation analysis, including genes not previously identified in CBD. A multiomic model of top DNA methylation and gene expression features demonstrated that the first component separated CBD from other samples and the second component separated control subjects from remaining samples. The top features on component one were enriched for T-lymphocyte function, and the top features on component two were enriched for innate immune signaling. We identified six differentially abundant proteins in CBD versus BeS, with two (SIT1 and SH2D1A) selected as important RNA features in the multiomic model. Our integrated analysis of DNA methylation, gene expression, and proteins in the lung identified multiomic signatures of CBD that differentiated it from BeS and control subjects.

Rationale and design of the SARCoidosis Outcomes in all respiratory Viral Infectious Diseases (SARCOVID) Study.

INTRODUCTION: Respiratory infections are ubiquitous. The COVID-19 pandemic has refocused our attention on how morbid and potentially fatal they can be, and how host factors have an impact on the clinical course and outcomes. Due to a range of vulnerabilities, patients with sarcoidosis may be at higher risk of poor outcomes from respiratory infections. The objective of the SARCoidosis Outcomes in all respiratory Viral Infectious Diseases (SARCOVID) Study is to determine the short-term and long-term impacts of respiratory viral illnesses (COVID-19 and non-COVID-19) in sarcoidosis. METHODS AND ANALYSIS: Up to 20 clinical sites across the USA are participating in the recruitment of 2000 patients for this observational, prospective study. To ensure that the study cohort is representative of the general population with sarcoidosis, participating sites include those dedicated to reaching under-represented minorities or patients from non-urban areas. Baseline data on demographic features, comorbidities, sarcoidosis characteristics and pre-enrolment lung function will be captured at study entry. During this 3-year study, all acute respiratory infectious events (from SARS-CoV-2 and any other respiratory pathogen) will be assessed and recorded at quarterly intervals. The level of required medical care and survival outcomes determine infection severity, and the impact of infection on quality of life measures will be recorded. Post-infection lung function and imaging results will measure the long-term impact on the trajectory of sarcoidosis. Patients will be analysed according to the clinical phenotypes of cardiac and fibrotic pulmonary sarcoidosis. Control groups include non-infected patients with sarcoidosis and patients with non-sarcoidosis interstitial lung disease. ETHICS AND DISSEMINATION: Each site received local institutional review board approval prior to enrolling patients, with the consent process determined by local institution standards. Data will be published in a timely manner (goal <12 months) at the conclusion of the 3-year follow-up period and will be made available upon request.

Realistic biomarkers from plasma extracellular vesicles for detection of beryllium exposure.

PURPOSE: Exposures related to beryllium (Be) are an enduring concern among workers in the nuclear weapons and other high-tech industries, calling for regular and rigorous biological monitoring. Conventional biomonitoring of Be in urine is not informative of cumulative exposure nor health outcomes. Biomarkers of exposure to Be based on non-invasive biomonitoring could help refine disease risk assessment. In a cohort of workers with Be exposure, we employed blood plasma extracellular vesicles (EVs) to discover novel biomarkers of exposure to Be. METHODS: EVs were isolated from plasma using size-exclusion chromatography and subjected to mass spectrometry-based proteomics. A protein-based classifier was developed using LASSO regression and validated by ELISA. RESULTS: We discovered a dual biomarker signature comprising zymogen granule protein 16B and putative protein FAM10A4 that differentiated between Be-exposed and -unexposed subjects. ELISA-based quantification of the biomarkers in an independent cohort of samples confirmed higher expression of the signature in the Be-exposed group, displaying high predictive accuracy (AUROC = 0.919). Furthermore, the biomarkers efficiently discriminated high- and low-exposure groups (AUROC = 0.749). CONCLUSIONS: This is the first report of EV biomarkers associated with Be exposure and exposure levels. The biomarkers could be implemented in resource-limited settings for Be exposure assessment.

Sex-Specific Differences in MicroRNA Expression During Human Fetal Lung Development.

Background: Sex-specific differences in fetal lung maturation have been well described; however, little is known about the sex-specific differences in microRNA (miRNA) expression during human fetal lung development. Interestingly, many adult chronic lung diseases also demonstrate sex-specific differences in prevalence. The developmental origins of health and disease hypothesis suggests that these sex-specific differences in fetal lung development may influence disease susceptibility later in life. In this study, we performed miRNA sequencing on human fetal lung tissue samples to investigate differential expression of miRNAs between males and females in the pseudoglandular stage of lung development. We hypothesized that differences in miRNA expression are present between sexes in early human lung development and may contribute to the sex-specific differences seen in pulmonary diseases later in life. Methods: RNA was isolated from human fetal lung tissue samples for miRNA sequencing. The count of each miRNA was modeled by sex using negative binomial regression models in DESeq2, adjusting for post-conception age, age2, smoke exposure, batch, and RUV factors. We tested for differential expression of miRNAs by sex, and for the presence of sex-by-age interactions to determine if miRNA expression levels by age were distinct between males and females. Results: miRNA expression profiles were generated on 298 samples (166 males and 132 females). Of the 809 miRNAs expressed in human fetal lung tissue during the pseudoglandular stage of lung development, we identified 93 autosomal miRNAs that were significantly differentially expressed by sex and 129 miRNAs with a sex-specific pattern of miRNA expression across the course of the pseudoglandular period. Conclusion: Our study demonstrates differential expression of numerous autosomal miRNAs between the male and female developing human lung. Additionally, the expression of some miRNAs are modified by age across the pseudoglandular stage in a sex-specific way. Some of these differences in miRNA expression may impact susceptibility to pulmonary disease later in life. Our results suggest that sex-specific miRNA expression during human lung development may be a potential mechanism to explain sex-specific differences in lung development and may impact subsequent disease susceptibility.

Clinical phenotyping in sarcoidosis using cluster analysis.

BACKGROUND: Most phenotyping paradigms in sarcoidosis are based on expert opinion; however, no paradigm has been widely adopted because of the subjectivity in classification. We hypothesized that cluster analysis could be performed on common clinical variables to define more objective sarcoidosis phenotypes. METHODS: We performed a retrospective cohort study of 554 sarcoidosis cases to identify distinct phenotypes of sarcoidosis based on 29 clinical features. Model-based clustering was performed using the VarSelLCM R package and the Integrated Completed Likelihood (ICL) criteria were used to estimate number of clusters. To identify features associated with cluster membership, features were ranked based on variable importance scores from the VarSelLCM model, and additional univariate tests (Fisher's exact test and one-way ANOVA) were performed using q-values correcting for multiple testing. The Wasfi severity score was also compared between clusters. RESULTS: Cluster analysis resulted in 6 sarcoidosis phenotypes. Salient characteristics for each cluster are as follows: Phenotype (1) supranormal lung function and majority Scadding stage 2/3; phenotype (2) supranormal lung function and majority Scadding stage 0/1; phenotype (3) normal lung function and split Scadding stages between 0/1 and 2/3; phenotype (4) obstructive lung function and majority Scadding stage 2/3; phenotype (5) restrictive lung function and majority Scadding stage 2/3; phenotype (6) mixed obstructive and restrictive lung function and mostly Scadding stage 4. Although there were differences in the percentages, all Scadding stages were encompassed by all of the phenotypes, except for phenotype 1, in which none were Scadding stage 4. Clusters 4, 5, 6 were significantly more likely to have ever been on immunosuppressive treatment and had higher Wasfi disease severity scores. CONCLUSIONS: Cluster analysis produced 6 sarcoidosis phenotypes that demonstrated less severe and severe phenotypes. Phenotypes 1, 2, 3 have less lung function abnormalities, a lower percentage on immunosuppressive treatment and lower Wasfi severity scores. Phenotypes 4, 5, 6 were characterized by lung function abnormalities, more parenchymal abnormalities, an increased percentage on immunosuppressive treatment and higher Wasfi severity scores. These data support using cluster analysis as an objective and clinically useful way to phenotype sarcoidosis subjects and to empower clinicians to identify those with more severe disease versus those who have less severe disease, independent of Scadding stage.

Cluster activation mapping with application to computed tomography scans of the lung.

Purpose: An open question in deep clustering is how to explain what in the image is driving the cluster assignments. This is especially important for applications in medical imaging when the derived cluster assignments may inform decision-making or create new disease subtypes. We develop cluster activation mapping (CLAM), which is methodology to create localization maps highlighting the image regions important for cluster assignment. Approach: Our approach uses a linear combination of the activation channels from the last layer of the encoder within a pretrained autoencoder. The activation channels are weighted by a channelwise confidence measure, which is a modification of score-CAM. Results: Our approach performs well under medical imaging-based simulation experiments, when the image clusters differ based on size, location, and intensity of abnormalities. Under simulation, the cluster assignments were predicted with 100% accuracy when the number of clusters was set at the true value. In addition, applied to computed tomography scans from a sarcoidosis population, CLAM identified two subtypes of sarcoidosis based purely on CT scan presentation, which were significantly associated with pulmonary function tests and visual assessment scores, such as ground-glass, fibrosis, and honeycombing. Conclusions: CLAM is a transparent methodology for identifying explainable groupings of medical imaging data. As deep learning networks are often criticized and not trusted due to their lack of interpretability, our contribution of CLAM to deep clustering architectures is critical to our understanding of cluster assignments, which can ultimately lead to new subtypes of diseases.