Assessing the inter-observer variability of Computer-Aided Nodule Assessment and Risk Yield (CANARY) to characterize lung adenocarcinomas

A computed tomography-based score indicative of lung cancer aggression (SILA) predicts lung adenocarcinomas with low malignant potential or vascular invasion

BACKGROUND

Histologic grading of lung adenocarcinoma (LUAD) is predictive of outcome but is only possible after surgical resection. A radiomic biomarker predictive of grade has the potential to improve preoperative management of early-stage LUAD.

OBJECTIVE

Validate a prognostic radiomic score indicative of lung cancer aggression (SILA) in surgically resected stage I LUAD (n= 161) histologically graded as indolent low malignant potential (LMP), intermediate, or aggressive vascular invasive (VI) subtypes.

METHODS

The SILA scores were generated from preoperative CT-scans using the previously validated Computer-Aided Nodule Assessment and Risk Yield (CANARY) software.

RESULTS

Cox proportional regression showed significant association between the SILA and 7-year recurrence-free survival (RFS) in a univariate (p< 0.05) and multivariate (p< 0.05) model incorporating age, gender, smoking status, pack years, and extent of resection. The SILA was positively correlated with invasive size (spearman r= 0.54, p= 8.0 × 10 - 14) and negatively correlated with percentage of lepidic histology (spearman r=-0.46, p= 7.1 × 10 - 10). The SILA predicted indolent LMP with an area under the receiver operating characteristic (ROC) curve (AUC) of 0.74 and aggressive VI with an AUC of 0.71, the latter remaining significant when invasive size was included as a covariate in a logistic regression model (p< 0.01).

CONCLUSIONS

The SILA scoring of preoperative CT scans was prognostic and predictive of resected pathologic grade.

Assessment of interobserver concordance in radiomic tools for lung nodule classification, with a focus on BRODERS and SILA

While CT lung cancer screening reduces lung cancer-specific mortality, there are remaining challenges. Radiomic tools promiss to address these challenges, however, they are subject to interobserver variability if semi-automated segmentation techniques are used. Herein we report interobserver variability for two validated radiomic tools, BRODERS (Benign versus aggRessive nODule Evaluation using Radiomic Stratification) and CANARY (Computer-Aided Nodule Assessment and Risk Yield). We retrospectively analyzed the CT images of 95 malignant lung nodules of the adenocarcinoma spectrum using BRODERS and CANARY. Cases were identified at Mayo Clinic (n = 45) and Vanderbilt University Medical Center and Nashville/Veteran Administration Tennessee Valley Health Care System (n = 50). Three observers with different training levels (medical student, internal medicine resident and thoracic radiology fellow) each performed lung nodule segmentation. All methods were carried out in accordance with relevant guidelines and regulations. Interclass correlation coefficients (ICC) of 0.77, 0.98 and 0.97 for the average nodule volume, BRODERS cancer probability and Score Indicative of Lesion Aggression (SILA) which summarizes the distribution of the CANARY exemplars indicated good to excellent reliability, respectively. The dice similarity coefficient was 0.79 and 0.81 for the data sets from the two institutions. BRODERS and CANARY are robust radiomics tools with excellent interobserver variability. These tools are simple and reliable regardless the observer/operator's level of training.

The impact of the lung EDRN-CVC on Phase 1, 2, & 3 biomarker validation studies

The Early Detection Research Network’s (EDRN) purpose is to discover, develop and validate biomarkers and imaging methods to detect early-stage cancers or at-risk individuals. The EDRN is composed of sites that fall into four categories: Biomarker Developmental Laboratories (BDL), Biomarker Reference Laboratories (BRL), Clinical Validation Centers (CVC) and Data Management and Coordinating Centers. Each component has a crucial role to play within the mission of the EDRN. The primary role of the CVCs is to support biomarker developers through validation trials on promising biomarkers discovered by both EDRN and non-EDRN investigators. The second round of funding for the EDRN Lung CVC at Vanderbilt University Medical Center (VUMC) was funded in October 2016 and we intended to accomplish the three missions of the CVCs: To conduct innovative research on the validation of candidate biomarkers for early cancer detection and risk assessment of lung cancer in an observational study; to compare biomarker performance; and to serve as a resource center for collaborative research within the Network and partner with established EDRN BDLs and BRLs, new laboratories and industry partners. This report outlines the impact of the VUMC EDRN Lung CVC and describes the role in promoting and validating biological and imaging biomarkers.

Radiomics/Radiogenomics in Lung Cancer: Basic Principles and Initial Clinical Results

Simple Summary

Radiogenomics is a promising new approach in cancer assessment, providing an evaluation of the molecular basis of imaging phenotypes after establishing associations between radiological features and molecular features at the genomic–transcriptomic–proteomic level. This review focuses on describing key aspects of radiogenomics while discussing limitations of translatability to the clinic and possible solutions to these challenges, providing the clinician with an up-to-date handbook on how to use this new tool.

Abstract

Lung cancer is the leading cause of cancer-related deaths worldwide, and elucidation of its complicated pathobiology has been traditionally targeted by studies incorporating genomic as well other high-throughput approaches. Recently, a collection of methods used for cancer imaging, supplemented by quantitative aspects leading towards imaging biomarker assessment termed “radiomics”, has introduced a novel dimension in cancer research. Integration of genomics and radiomics approaches, where identifying the biological basis of imaging phenotypes is feasible due to the establishment of associations between molecular features at the genomic–transcriptomic–proteomic level and radiological features, has recently emerged termed radiogenomics. This review article aims to briefly describe the main aspects of radiogenomics, while discussing its basic limitations related to lung cancer clinical applications for clinicians, researchers and patients.

On the Automation of Radiomics-Based Identification and Characterization of NSCLC

Proper detection and accurate characterization of Non-Small Cell Lung Cancer (NSCLC) are an open challenge in the imaging field. Biomedical imaging is fundamental in lung cancer assessment and offers the possibility of calculating predictive biomarkers impacting patients' management. Within this context, radiomics, which consists of extracting quantitative features from digital images, shows encouraging results for clinical applications, but the sub-optimal standardization of the procedure and the lack of definitive results are still a concern in the field. For these reasons, this work proposes the design and development of LuCIFEx, a fully-automated pipeline for non-invasive in-vivo characterization of NSCLC, aiming to speed up the analysis process and enable an early diagnosis of the tumor.LuCIFEx pipeline relies on routinely acquired [18F]FDG-PET/CT images for the automatic segmentation of the cancer lesion, allowing the computation of accurate radiomic features, then employed for cancer characterization through Machine Learning algorithms. The proposed multi-stage segmentation process can identify the lesion with a mean accuracy of 94.2±5.0%. Finally, the proposed data analysis pipeline demonstrates the potential of PET/CT features for the automatic recognition of lung metastases and NSCLC histological subtypes, while highlighting the main current limitations of the radiomic approach.

Evaluation of Computer-Aided Nodule Assessment and Risk Yield (CANARY) in Korean patients for prediction of invasiveness of ground-glass opacity nodule

Differentiating the invasiveness of ground-glass nodules (GGN) is clinically important, and several institutions have attempted to develop their own solutions by using computed tomography images. The purpose of this study is to evaluate Computer-Aided Analysis of Risk Yield (CANARY), a validated virtual biopsy and risk-stratification machine-learning tool for lung adenocarcinomas, in a Korean patient population. To this end, a total of 380 GGNs from 360 patients who underwent pulmonary resection in a single institution were reviewed. Based on the Score Indicative of Lung Cancer Aggression (SILA), a quantitative indicator of CANARY analysis results, all of the GGNs were classified as “indolent” (atypical adenomatous hyperplasia, adenocarcinomas in situ, or minimally invasive adenocarcinoma) or “invasive” (invasive adenocarcinoma) and compared with the pathology reports. By considering the possibility of uneven class distribution, statistical analysis was performed on the 1) entire cohort and 2) randomly extracted six sets of class-balanced samples. For each trial, the optimal cutoff SILA was obtained from the receiver operating characteristic curve. The classification results were evaluated using several binary classification metrics. Of a total of 380 GGNs, the mean SILA for 65 (17.1%) indolent and 315 (82.9%) invasive lesions were 0.195±0.124 and 0.391±0.208 (p < 0.0001). The area under the curve (AUC) of each trial was 0.814 and 0.809, with an optimal threshold SILA of 0.229 for both. The macro F1-score and geometric mean were found to be 0.675 and 0.745 for the entire cohort, while both scored 0.741 in the class-equalized dataset. From these results, CANARY could be confirmed acceptable in classifying GGN for Korean patients after the cutoff SILA was calibrated. We found that adjusting the cutoff SILA is needed to use CANARY in other countries or races, and geometric mean could be more objective than F1-score or AUC in the binary classification of imbalanced data.

Validation of the BRODERS classifier (Benign versus aggRessive nODule Evaluation using Radiomic Stratification), a novel HRCT-based radiomic classifier for indeterminate pulmonary nodules

INTRODUCTION

Implementation of low-dose chest computed tomography (CT) lung cancer screening and the ever-increasing use of cross-sectional imaging are resulting in the identification of many screen- and incidentally detected indeterminate pulmonary nodules. While the management of nodules with low or high pre-test probability of malignancy is relatively straightforward, those with intermediate pre-test probability commonly require advanced imaging or biopsy. Noninvasive risk stratification tools are highly desirable.

METHODS

We previously developed the BRODERS classifier (Benign versus aggRessive nODule Evaluation using Radiomic Stratification), a conventional predictive radiomic model based on eight imaging features capturing nodule location, shape, size, texture and surface characteristics. Herein we report its external validation using a dataset of incidentally identified lung nodules (Vanderbilt University Lung Nodule Registry) in comparison to the Brock model. Area under the curve (AUC), as well as sensitivity, specificity, negative and positive predictive values were calculated.

RESULTS

For the entire Vanderbilt validation set (n=170, 54% malignant), the AUC was 0.87 (95% CI 0.81-0.92) for the Brock model and 0.90 (95% CI 0.85-0.94) for the BRODERS model. Using the optimal cut-off determined by Youden's index, the sensitivity was 92.3%, the specificity was 62.0%, the positive (PPV) and negative predictive values (NPV) were 73.7% and 87.5%, respectively. For nodules with intermediate pre-test probability of malignancy, Brock score of 5-65% (n=97), the sensitivity and specificity were 94% and 46%, respectively, the PPV was 78.4% and the NPV was 79.2%.

CONCLUSIONS

The BRODERS radiomic predictive model performs well on an independent dataset and may facilitate the management of indeterminate pulmonary nodules.

Computed Tomography-Based Score Indicative of Lung Cancer Aggression (SILA) Predicts the Degree of Histologic Tissue Invasion and Patient Survival in Lung Adenocarcinoma Spectrum

OBJECTIVE

Most computed tomography (CT)-detected lung cancers are adenocarcinomas (ACs), representing lesions with variable tissue invasion, aggressiveness, and clinical outcome. Visual radiologic characterization of AC pulmonary nodules is both inconsistent and inadequate to confidently predict histopathologic classification or prognosis. Comprehensive pathologic interpretation requires full nodule resection. We have described a computerized scoring system for AC detected on CT scans that can noninvasively estimate the degree of histologic invasion and simultaneously predict patient survival.

METHODS

The Computer-Aided Nodule Assessment and Risk Yield has been validated to characterize CT-detected nodules across the spectrum of AC. With the use of unsupervised clustering, nine natural exemplars were identified as basic radiographic features of AC nodules. We now introduce the Score Indicative of Lung Cancer Aggression (SILA), which is a cumulative aggregate of normalized distributions of ordered Computer-Aided Nodule Assessment and Risk Yield exemplars. The SILA values for each of 237 unique nodules in AC were compared with the histopathologically defined maximum linear extent of tumor invasion. With use of the SILA, Kaplan-Meier survival and Cox proportionality analysis were performed on patients with stage I AC, who comprised a subset of our cohort.

RESULTS

The SILA discriminated between indolent and invasive AC (p < 0.0001). In addition, prediction of linear extent of histopathologic tumor invasion was possible. In stage I AC, three separate SILA prognosis groups were identified: indolent, intermediate, and poor, with 5-year survival rates of 100%, 79%, 58%, respectively. Cox proportionality hazard modeling predicted a 50% increase in mortality, for a 0.1 unit increase in the SILA over a median follow-up time of 3.6 years (p < 0.0002).

CONCLUSIONS

The SILA is a computer-based analytic measure allowing noninvasive approximation of histologic invasion and prediction of patient survival in CT-detected AC nodules.

Lung Cancer Detection using Co-learning from Chest CT Images and Clinical Demographics

Early detection of lung cancer is essential in reducing mortality. Recent studies have demonstrated the clinical utility of low-dose computed tomography (CT) to detect lung cancer among individuals selected based on very limited clinical information. However, this strategy yields high false positive rates, which can lead to unnecessary and potentially harmful procedures. To address such challenges, we established a pipeline that co-learns from detailed clinical demographics and 3D CT images. Toward this end, we leveraged data from the Consortium for Molecular and Cellular Characterization of Screen-Detected Lesions (MCL), which focuses on early detection of lung cancer. A 3D attention-based deep convolutional neural net (DCNN) is proposed to identify lung cancer from the chest CT scan without prior anatomical location of the suspicious nodule. To improve upon the non-invasive discrimination between benign and malignant, we applied a random forest classifier to a dataset integrating clinical information to imaging data. The results show that the AUC obtained from clinical demographics alone was 0.635 while the attention network alone reached an accuracy of 0.687. In contrast when applying our proposed pipeline integrating clinical and imaging variables, we reached an AUC of 0.787 on the testing dataset. The proposed network both efficiently captures anatomical information for classification and also generates attention maps that explain the features that drive performance.

Computer Aided Nodule Analysis and Risk Yield (CANARY) characterization of adenocarcinoma: radiologic biopsy, risk stratification and future directions

The majority of incidentally and screen-detected lung cancers are adenocarcinomas. Optimal management of these tumors is clinically challenging due to variability in tumor histopathology and behavior. Invasive adenocarcinoma (IA) is generally aggressive while adenocarcinoma in situ (AIS) and minimally invasive adenocarcinoma (MIA) may be extremely indolent. Computer Aided Nodule Analysis and Risk Yield (CANARY) is a quantitative computed tomography (CT) analysis tool that allows non-invasive assessment of tumor characteristics. This analysis may obviate the need for tissue biopsy and facilitate the risk stratification of adenocarcinoma of the lung. CANARY was developed by unsupervised machine learning techniques using CT data of histopathologically-characterized adenocarcinomas of the lung. This technique identified 9 distinct exemplars that constitute the spectrum of CT features found in adenocarcinoma of the lung. The distributions of these features in a nodule correlate with histopathology. Further automated clustering of CANARY nodules defined three distinct groups that have distinctly different post-resection disease free survival (DFS). CANARY has been validated within the NLST cohort and multiple other cohorts. Using semi-automated segmentation as input to CANARY, there is excellent repeatability and interoperator correlation of results. Confirmation and longitudinal tracking of indolent adenocarcinoma with CANARY may ultimately add decision support in nuanced cases where surgery may not be in the best interest of the patient due to competing comorbidity. Currently under investigation is CANARY's role in detecting differing driver mutations and tumor response to targeted chemotherapeutics. Combining the results from CANARY analysis with clinical information and other quantitative techniques such as analysis of the tumor-free surrounding lung may aid in building more powerful predictive models. The next step in CANARY investigation will be its prospective application, both in selecting low-risk stage 1 adenocarcinoma for active surveillance and investigation in selecting high-risk early stage adenocarcinoma for adjuvant therapy.