Remotely operated robotic microscopy for rapid diagnosis of bronchoscopic cytology specimens

BACKGROUND

Obtaining a diagnosis and treating pulmonary malignancies during the same anesthesia requires either an on-site pathologist or a system for remotely evaluating microscopic images. Cytology specimens are challenging to remotely assess given the need to navigate through dispersed and three-dimensional cell clusters. Remote navigation is possible using robotic telepathology, but data are limited on the ease of use of current systems, particularly for pulmonary cytology.

METHODS

Air dried modified Wright-Giemsa stained slides from 26 touch preparations of transbronchial biopsies and 27 smears of endobronchial ultrasound guided fine needle aspirations were scored for ease of adequacy assessment and ease of diagnosis on robotic (rmtConnect Microscope) and non-robotic telecytology platforms. Diagnostic classifications were compared between glass slides and the robotic and non-robotic telecytology assessments.

RESULTS

Compared to non-robotic telecytology, robotic telecytology had a greater ease of adequacy assessment and non-inferior ease of diagnosis. The median time to diagnosis using robotic telecytology was 85 s (range 28-190 s). Diagnostic categories were concordant for 76% of cases in robotic versus non-robotic telecytology and 78% of cases in robotic telecytology versus glass slide diagnosis. Weighted Cohen's kappa scores for agreement in these comparisons were 0.84 and 0.72, respectively.

CONCLUSIONS

Use of a remote-controlled robotic microscope improved the ease of adequacy assessment compared to non-robotic telecytology and enabled strongly concordant diagnoses to be expediently rendered. This study provides evidence that modern robotic telecytology is a feasible and user-friendly method of remotely and potentially intraoperatively rendering adequacy assessments and diagnoses on bronchoscopic cytology specimens.

Uncertainty-informed deep learning models enable high-confidence predictions for digital histopathology

A model's ability to express its own predictive uncertainty is an essential attribute for maintaining clinical user confidence as computational biomarkers are deployed into real-world medical settings. In the domain of cancer digital histopathology, we describe a clinically-oriented approach to uncertainty quantification for whole-slide images, estimating uncertainty using dropout and calculating thresholds on training data to establish cutoffs for low- and high-confidence predictions. We train models to identify lung adenocarcinoma vs. squamous cell carcinoma and show that high-confidence predictions outperform predictions without uncertainty, in both cross-validation and testing on two large external datasets spanning multiple institutions. Our testing strategy closely approximates real-world application, with predictions generated on unsupervised, unannotated slides using predetermined thresholds. Furthermore, we show that uncertainty thresholding remains reliable in the setting of domain shift, with accurate high-confidence predictions of adenocarcinoma vs. squamous cell carcinoma for out-of-distribution, non-lung cancer cohorts.

The use of artificial intelligence with uncertainty estimation to predict lung cancer relapse from histopathology

8549

Background: First-line treatment of early stage non-small cell lung cancer (NSCLC) is surgical resection, but with a 5-year survival of only 54%, rates of future relapse are high. Identifying patients at high risk of relapse can help guide adjuvant treatment decisions. Deep convolutional neural network (DCNN) AI models trained on tumor histology have shown incredible flexibility as potential biomarkers, both in lung cancer and more generally across many malignancies. While DCNN models can obtain extremely accurate results when used for routine purposes such as diagnosis, subtyping, and grading of malignancies, most models trained for prognostication or treatment response prediction do not reach performance sufficient for clinical application. Uncertainty quantification (UQ) – a family of techniques that give DCNN models the ability to report confidence alongside predictions – is an underexplored avenue in cancer AI that may help further improve performance and clinical application of models designed to provide clinicians with estimations of risk. Methods: To explore the potential use of UQ in clinically-oriented DCNN models, we trained models on a single-institution, retrospective digital tumor histology slide cohort from patients with Stage I-III NSCLC who underwent surgical resection to predict risk of future relapse. Estimation of uncertainty was performed using dropout as a Bayesian approximation, and uncertainty thresholds were calculated from training sets using a novel method to identify and remove low-confidence predictions. For comparison, a separate multivariate logistic regression was trained in cross-validation using known clinical risk factors. Results: We trained DCNN models on slides from 198 patients (40 relapsed, 158 without relapse). In this cohort, 130 patients had stage I disease (65.6%), 42 had stage II (21.2%), and 26 had stage III (13.2%). The average age was 69 years, 85% were current or previous smokers, and 38 received guideline-concordant adjuvant chemotherapy. Without UQ, a DCNN model predicted risk of future relapse with an average area under receiver operator curve (AUROC) of 0.74 across three-fold cross-validation. Using UQ estimation, 63% of slides were reported with high confidence. Relapse prediction was significantly improved in the high-confidence cohort, with an average AUROC of 0.83 in cross-validation. With a specificity of 70%, this corresponds to an average sensitivity of 86.8% across the three cross-folds (79.1%, 85.5%, and 95.7%). In comparison, a clinical-only multivariate regression model predicted relapse with a cross-validated AUROC of 0.67. Conclusions: This method of uncertainty quantification appears to be a powerful tool to predict lung cancer recurrence risk from digital histopathology while simultaneously providing clinicians with a measurement of algorithm trustworthiness.

Optimizing clinical cytology touch preparations for next generation sequencing

Intraoperative diagnosis is routinely performed on cytology touch preparations (TPs) from core needle biopsies (CNBs). Current interest promotes their utility as an important source of patient tissue for clinical genomic testing. Herein we present whole genome structural variant analysis (SVA) from mate-pair sequencing (MPseq) and whole exome sequencing (WES) mutation calling in DNA directly whole genome amplified (WGA) from TPs. Chromosomal copy changes and somatic DNA junction detection from MPseq of TPs were highly consistent with associated CNBs and bulk resected tissues in all cases. While increased frequency coverage noise from limitations of amplification of limited sample input was significant, this was effectively compensated by natural tumor enrichment during the TP process, which also enhanced variant detection and loss of heterozygosity evaluations from WES. This novel TP methodology enables expanded utility of frequently limited CNB for both clinical and research genomic testing.

Neoantigenic Potential of Complex Chromosomal Rearrangements in Mesothelioma

INTRODUCTION

Malignant pleural mesothelioma is a disease primarily associated with exposure to the carcinogen asbestos. Whereas other carcinogen-related tumors are associated with a high tumor mutation burden, mesothelioma is not. We sought to resolve this discrepancy.

METHODS

We used mate-pair (n = 22), RNA (n = 28), and T cell receptor sequencing along with in silico predictions and immunologic assays to understand how structural variants of chromosomes affect the transcriptome.

RESULTS

We observed that inter- or intrachromosomal rearrangements were present in every specimen and were frequently in a pattern of chromoanagenesis such as chromoplexy or chromothripsis. Transcription of rearrangement-related junctions was predicted to result in many potential neoantigens, some of which were proven to bind patient-specific major histocompatibility complex molecules and to expand intratumoral T cell clones. T cells responsive to these predicted neoantigens were also present in a patient's circulating T cell repertoire. Analysis of genomic array data from the mesothelioma cohort in The Cancer Genome Atlas suggested that multiple chromothriptic-like events negatively impact survival.

CONCLUSIONS

Our findings represent the discovery of potential neoantigen expression driven by structural chromosomal rearrangements. These results may have implications for the development of novel immunotherapeutic strategies and the selection of patients to receive immunotherapies.

Abstract 5726: Rearrangement-related peptides with neoantigenic potential in malignant pleural mesothelioma

Malignant pleural mesothelioma is a disease primarily associated with exposure to the carcinogen asbestos. Consistent with this carcinogenic exposure, cytogenetic analyses have identified multiple recurrent structural chromosomal abnormalities in this malignancy, but more recent high-throughput sequencing evaluations of point mutations suggest that there is a low mutational burden in mesothelioma. Since tumor mutational burden has been correlated with responses to treatment with immune checkpoint inhibitors such as nivolumab, it was not consistent that patients with mesothelioma and low mutation burdens would have similar response rates in clinical trials with immune checkpoint inhibitors as patients with non-small cell lung cancer which is associated with a high mutation burden. In order to reconcile these differences, and given the potential for an improved understanding of the molecular pathogenesis of mesothelioma to improve therapeutic options, we used mate-pair sequencing (MPseq) and RNA sequencing (RNAseq) to understand how structural variants affect the transcriptome. MPseq differs from standard next generation sequencing approaches by tiling the whole genome with larger fragments (2-5kb) to reliably detect structural variants such as insertions, deletions and rearrangements. Amongst 22 mesothelioma specimens there were 1535 chromosomal rearrangements (median 41, range 3-298 per specimen), that resulted in junctions or novel fusions of non-coding DNA or genes. Six-hundred thirty-seven of these rearrangements (median 22, range 5-103 range per specimen) resulted in novel fusions of genes. Many of these inter- or intra-chromosomal rearrangements were consistent with a pattern of chromoanagesis such as chromoplexy or chromothripsis. Chromosomal rearrangements detected by MPseq were used to guide analysis of RNAseq data and revealed that these chromosomal junctions resulted in the expression of 179 novel amino acid sequences (median 5, 0-51 range per specimen). To determine whether transcription of chromosomal rearrangement-related junctions have neoantigenic potential, we used in silico tools to determine whether any of the expressed junctions contained peptides that could be presented by patient-specific HLA molecules. The top candidate rearrangement-related peptides with neoantigenic potential bound patient-specific HLA molecules nearly as well or as well as a positive control in competitive binding assays. Our findings represent the discovery of potential neoantigen expression driven by structural chromosomal rearrangements. These results may have implications for the development of novel therapeutic strategies, the selection of patients to receive immunotherapy, and blood-based treatment monitoring strategies.

Citation Format: Aaron S. Mansfield, Tobias Peikert, James B. Smadbeck, Julia B. Udell, Farhad Kosari, Stephen J. Murphy, Hongzheng Ren, Vishnu V. Serla, Janet L. Schaefer Klein, Giannoula Karagouga, Faye R. Harris, Carlos Sosa, Sarah H. Johnson, Wendy Nevala, Svetomir N. Markovic, Aaron O. Bungum, Eric S. Edell, Haidong Dong, John C. Cheville, Marie Christine Aubry, Jin Jen, George Vasmatzis. Rearrangement-related peptides with neoantigenic potential in malignant pleural mesothelioma [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2018; 2018 Apr 14-18; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2018;78(13 Suppl):Abstract nr 5726.

ANK1 Methylation regulates expression of MicroRNA-486-5p and discriminates lung tumors by histology and smoking status

The intragenic tumor-suppressor microRNA miR-486-5p is often down-regulated in non-small cell lung cancer (NSCLC) but the mechanism is unclear. This study investigated epigenetic co-regulation of miR-486-5p and its host gene ANK1. MiR-486-5p expression in lung tumors and cell lines was significantly reduced compared to normal lung (p < 0.001) and is strongly correlated with ANK1 expression. In vitro, siRNA-mediated ANK1 knockdown in NSCLC cells also reduced miR-486-5p while the DNA methylation inhibitor 5-aza-2'-deoxycytidine induced expression of both. ANK1 promoter CpG island was unmethylated in normal lung but methylated in 45% (118/262) lung tumors and 55% (17/31) NSCLC cell lines. After adjustment for tumor histology and smoking, methylation was significantly more prevalent in adenocarcinoma (101/200, 51%) compared to squamous cell carcinoma (17/62, 27%), p < 0.001; HR = 3.513 (CI: 1.818-6.788); and in smokers (73/128, 57%) than never-smokers (28/72, 39%), p = 0.014; HR = 2.086 (CI: 1.157-3.759). These results were independently validated using quantitative methylation data for 809 NSCLC cases from The Cancer Genome Atlas project. Together, our data indicate that aberrant ANK1 methylation is highly prevalent in lung cancer, discriminate tumors by histology and patients' smoking history, and contributes to miR-486-5p repression.

Blood-based lung cancer biomarkers identified through proteomic discovery in cancer tissues, cell lines and conditioned medium

Background

Support for early detection of lung cancer has emerged from the National Lung Screening Trial (NLST), in which low-dose computed tomography (LDCT) screening reduced lung cancer mortality by 20 % relative to chest x-ray. The US Preventive Services Task Force (USPSTF) recently recommended annual screening for the high-risk population, concluding that the benefits (life years gained) outweighed harms (false positive findings, abortive biopsy/surgery, radiation exposure). In making their recommendation, the USPSTF noted that the moderate net benefit of screening was dependent on the resolution of most false-positive results without invasive procedures. Circulating biomarkers may serve as a valuable adjunctive tool to imaging.

Results

We developed a broad-based proteomics discovery program, integrating liquid chromatography/mass spectrometry (LC/MS) analyses of freshly resected lung tumor specimens (n = 13), lung cancer cell lines (n = 17), and conditioned media collected from tumor cell lines (n = 7). To enrich for biomarkers likely to be found at elevated levels in the peripheral circulation of lung cancer patients, proteins were prioritized based on predicted subcellular localization (secreted, cell-membrane associated) and differential expression in disease samples. 179 candidate biomarkers were identified. Several markers selected for further validation showed elevated levels in serum collected from subjects with stage I NSCLC (n = 94), relative to healthy smoker controls (n = 189). An 8-marker model was developed (TFPI, MDK, OPN, MMP2, TIMP1, CEA, CYFRA 21–1, SCC) which accurately distinguished subjects with lung cancer (n = 50) from high risk smokers (n = 50) in an independent validation study (AUC = 0.775).

Conclusions

Integrating biomarker discovery from multiple sample types (fresh tissue, cell lines and conditioned medium) has resulted in a diverse repertoire of candidate biomarkers. This unique collection of biomarkers may have clinical utility in lung cancer detection and diagnoses.

Electronic supplementary material

The online version of this article (doi:10.1186/s12014-015-9090-9) contains supplementary material, which is available to authorized users.

Abstract 5060: Gene network predicts overall survival in patients with primary lung adenocarcinoma

Lung adenocarcinoma is the most common type of primary lung cancer. Previous studies have shown that gene expression signatures may predict clinical outcome. In this study, we used a systems biology approach, weighted gene co-expression network analysis (WGCNA), to identify survival related networks and gene signatures. We first analyzed whole genome gene expression profiles in 82 lung adenocarcinoma tumor tissues from never smokers (smoked &lt;100 cigarettes lifetime) and found a significant correlation of an expression module with tumor grade of differentiation (r=0.38, p=4×10−4) and overall survival (OS, p =0.027, HR, 1.71; 95% CI, 1.06-2.74) after adjusting for potential confounders (age and tumor grade). When this module signature was incorporated and compared to the conventional model consisting of clinical variables only, the survival prediction accuracy was increased from 0.70 to 0.75. Gene Ontology (GO) enrichment analysis indicated this module was significantly enriched in biological process GO term “cell cycle” (p = 2.1×10−53). To explore the effect of this network module at germline level, we performed the WGCNA on normal lung tissues derived from 78 of the 82 patients and observed significant association of an expression module with OS (adjusted p=3×10−4, HR=1.98; 95% CI, 1.36-2.87). Interestingly, this module was also significantly enriched in cell cycle-related genes (p=4.1×10−41). To further validate this result, we analyzed an independent microarray gene expression dataset including 442 lung adenocarcinoma patients from NCI Director's Challenge Consortium. We found the same association: the cell cycle-enriched module was correlated with tumor grade of cell differentiation (r=0.55, p-value=5×10−37) and OS (adjusted p=7.4×10−3; HR, 1.20; 95% CI, 1.05-1.37). We further identified 30 genes in the OS-related module that overlapped between the discovery and validation data sets. Furthermore, all these 30 genes were found to be up-regulated in adenocarcinoma compared to adjacent normal lung tissues in the discovery data set. Of the 30 genes, three key genes, UBE2C, TPX2 and MELK, were identified with highest connectivity in cell cycle-enriched module; their expression levels were significantly associated with OS in both discovery and validation sets. UBE2C and TPX2 have been reported as gene signatures for human lung cancer prognosis in vitro lung carcinogenesis system; MELK was identified as a key regulator of the proliferation of malignant brain tumors and malignant grade in human astrocytomas, and was also associated with breast cancer prognosis. Our results suggest that genes involved in cell-cycle are likely related with tumor aggressiveness and therefore can predict survival in lung adenocarcinoma. Further studies are needed to validate the key genes that could improve survival prediction and provide potential new targets to intervene progression of lung adenocarcinoma.

Citation Format: {Authors}. {Abstract title} [abstract]. In: Proceedings of the 102nd Annual Meeting of the American Association for Cancer Research; 2011 Apr 2-6; Orlando, FL. Philadelphia (PA): AACR; Cancer Res 2011;71(8 Suppl):Abstract nr 5060. doi:10.1158/1538-7445.AM2011-5060

CT screening for lung cancer: five-year prospective experience

PURPOSE

To report results of a 5-year prospective low-dose helical chest computed tomographic (CT) study of a cohort at high risk for lung cancer.

MATERIALS AND METHODS

After informed written consent was obtained, 1520 individuals were enrolled. Protocol was approved by institutional review board and National Cancer Institute and was compliant with Health Insurance Portability and Accountability Act, or HIPAA. Participants were aged 50 years and older and had smoked for more than 20 pack-years. Participants underwent five annual (one initial and four subsequent) CT examinations. A significant downward shift was evaluated in non-small cell lung cancers detected initially from advanced stage down to stage I by using a one-sided binomial test of proportions. Poisson regression and Fisher exact tests were used for comparisons with Mayo Lung Project.

RESULTS

In 788 (52%) men and 732 (48%) women, 61% (927 of 1520) were current smokers, and 39% were former smokers. After five annual CT examinations, 3356 uncalcified lung nodules were identified in 1118 (74%) participants. Sixty-eight lung cancers were diagnosed (31 initial, 34 subsequent, three interval cancers) in 66 participants. Twenty-eight subsequent cases of non-small cell cancers were detected, of which 17 (61%; 95% confidence interval: 41%, 79%) were stage I tumors. Diameter of cancers detected subsequently was 5-50 mm (mean, 14.4 mm; median, 10.0 mm). Analysis for a more than 50% shift in proportion of stage I non-small cell cancer detection did not show statistical significance. Forty-eight participants died of various causes since enrollment. Lung cancer mortality rate for incidence portion of trial was 1.6 per 1000 person-years. There was no significant difference in lung cancer mortality rates of cancers detected in subsequent examinations between this trial and Mayo Lung Project after separation of participants into subsets (2.8 vs 2.0 per 1000 person-years, P = .43).

CONCLUSION

CT allows detection of early-stage lung cancers. Benign nodule detection rate is high. Results suggest no stage shift.