Download This Paper
Preprints with The Lancet is a collaboration between The Lancet Group of journals and SSRN to facilitate the open sharing of preprints for early engagement, community comment, and collaboration. Preprints available here are not Lancet publications or necessarily under review with a Lancet journal. These preprints are early-stage research papers that have not been peer-reviewed. The usual SSRN checks and a Lancet-specific check for appropriateness and transparency have been applied. The findings should not be used for clinical or public health decision-making or presented without highlighting these facts. For more information, please see the FAQs.
Lung Graph-Based Machine Learning for Identification of Fibrotic Interstitial Lung Diseases
22 Pages Posted: 18 Jul 2022
More...Abstract
Background: Fibrotic interstitial lung disease (f-ILD), represented by idiopathic pulmonary fibrosis, has become a major public health problem. Early detection is conducive to early intervention. However, early identification of f-ILD still a big challenge. Our aim was to develop and assess a lung graph-based machine learning method to identify f-ILD.
Methods: A total of 417 HRCT from 279 patients with confirmed ILD (f-ILD: 156 patients, 223 CT scans; non-fibrotic ILD: 123 patients, 194 CT scans) were included in this study. Among them, 80% of patients were grouped into training set and the remaining 20% were grouped in testing set. A lung graph-based machine learning model based on HRCT was developed for f-ILD diagnosis. In this approach, local radiomics features were extracted from an automatically generated geometric atlas of the lungs and used to build a series of specific lung graph models. Encoding these lung graphs, a lung descriptor was gained and became as a characterization of global radiomics feature distribution to diagnose f-ILD. Five-fold cross validation was used for selecting the optimal machine learning method. To assess the performance of the model, the area under the receiver operating characteristic curve (AUC), sensitivity, specificity, the positive predictive value and the negative predictive value were selected as the metrics. Meanwhile, it was compared with the performance of 3 radiologists.
Findings: A lung graph-based machine learning model achieved good performance at patient level across these five data splits (AUC= 0.973±0.019). The diagnostic performance of the model was further demonstrated in an external validation cohort with an accuracy of 0.986 for f-ILD identification, which was higher than three radiologists (radiologist M, 0.918; radiologist L, 0.822; radiologist B, 0.904). The AUC values were statistically significantly different between the model and radiologists A, B and C with p values of 0.0111, <0.0001 and <0.0038, respectively.
Interpretation: The lung graph-based machine learning model could identify f-ILD, and the diagnostic performance exceeded radiologists which could aid clinician to assess ILD objectively.
Funding Statement: This work was supported by This work was supported by National Key R & D Program of China (Nos. 2021YFC2500700 and 2016YFC0901101 to H.D) and the National Natural Science Foundation of China (No. 81870056 to H.D).
Declaration of Interests: The authors have no conflicts of interest to declare.
Ethics Approval Statement: This study was approved by the China-Japan Friendship Hospital Committee (2022-KY-031).
Keywords: fibrotic interstitial lung disease (f-ILD), machine learning, lung graph, high-resolution computed tomography (HRCT)
Suggested Citation: Suggested Citation