OP0150 MACHINE LEARNING APPROACHES FOR RISK MODELLING IN INTERSTITIAL LUNG DISEASE ASSOCIATED WITH SYSTEMIC SCLEROSIS USING HIGH DIMENSIONAL IMAGE ANALYSIS

Background: The interstitial lung disease (ILD) associated with connective tissue diseases including systemic sclerosis (SSc) is heterogenous characterized by reduced survival of approximately 3 years (1). “Radiomics’’ a field research which describes the in-depth analysis tissues computational retrieval high-dimensional quantitative features from medical images (2). Our previous study suggested capacity radiomics to differentiate between “high” and “low” risk groups for function decline in two independent cohorts (3). Objectives: •bTo develop robust, machine learning (ML) workflow “radiomics” data SSc-ILD select optimal methods prediction. •oTo predict time individual defined as relative ≥ 15% Forced Vital Capacity (FVC)% previously (3), using workflow. Methods: We investigated SSc-ILD: 90 patients (76.7% female, median age 57.5 years) University Hospital Zurich 66 (75.8% 61.0 Oslo Hospital’s. Patients were retrospectively selected if (3): a) diagnosed early/mild SSc according Very Early Diagnosis Systemic Sclerosis (VEDOSS) criteria, b) presence ILD on HRCT determined senior radiologist. For every subject, we 1,355 robust radiomic images. follow-up period was interval baseline visit last available visit. have developed systematic build predictive ML models. To reduce number redundant features, applied correlation thresholds. distinct 1) Lasso Penalized Regression feature selection, 2) Random Forest (RF) modeling R package ‘caret’. model, randomly divided derivation cohort into Training (70%) Holdout (30%) sets fivefold cross-validation (5kCV) classifier selection set only. Results: various features. Since model performance affected both, feature, assessed these factors first. Results filtering that combination threshold 0.9 regression proved best. As perform 5k CV workflow, present at least 2 entered optimization step. During RF classifier. detected positive actual predicted values Spearman’s rho = 0.313, p 0.167 0.341, 0.015 respectively, shown Figure 1. percentage variance remained modest both (Rsq 0.104) 0.126) datasets. Performance best, scatterplot decline. Conclusion: In summary, we: (1) allowed o methodology (i.e., selection), (2) provide models decline, significant values. References: [1]Hansell DM, Goldin JG, King TE, Jr., Lynch DA, Richeldi L, Wells AU. CT staging monitoring fibrotic clinical practice treatment trials: position paper Fleischner Society. Lancet Respir Med. 2015;3(6):483-96. [2]Lambin, P. et al. Radiomics: extracting more information advanced analysis. Eur. J. Cancer 48, 441–446 (2012). [3]Schniering Resolving phenotypic prognostic differences related computed tomography-based radiomics. https://www.medrxiv.org/content/10.1101/2020.06.09.20124800v1 Disclosure Interests: None declared