High-resolution computed tomography to differentiate chronic diffuse interstitial lung diseases with predominant ground-glass pattern using logical analysis of data.

OBJECTIVES: We evaluated the performance of high-resolution computed tomography (HRCT) to differentiate chronic diffuse interstitial lung diseases (CDILD) with predominant ground-glass pattern by using logical analysis of data (LAD). METHODS: A total of 162 patients were classified into seven categories: sarcoidosis (n = 38), connective tissue disease (n = 32), hypersensitivity pneumonitis (n = 18), drug-induced lung disease (n = 15), alveolar proteinosis (n = 12), idiopathic non-specific interstitial pneumonia (n = 10) and miscellaneous (n = 37). First, 40 CT attributes were investigated by the LAD to build up patterns characterising a category. From the association of patterns, LAD determined models specific to each CDILD. Second, data were recomputed by adding eight clinical attributes to the analysis. The 20 x 5 cross-folding method was used for validation. RESULTS: Models could be individualised for sarcoidosis, hypersensitivity pneumonitis, connective tissue disease and alveolar proteinosis. An additional model was individualised for drug-induced lung disease by adding clinical data. No model was demonstrated for idiopathic non-specific interstitial pneumonia and the miscellaneous category. The results showed that HRCT had a good sensitivity (>or=64%) and specificity (>or=78%) and a high negative predictive value (>or=93%) for diseases with a model. Higher sensitivity (>or=78%) and specificity (>or=89%) were achieved by adding clinical data. CONCLUSION: The diagnostic performance of HRCT is high and can be increased by adding clinical data.

Logical analysis of survival data: prognostic survival models by detecting high-degree interactions in right-censored data.

MOTIVATION: Survival analysis involves predicting the time to event for patients in a dataset, based on a set of recorded attributes. In this study we focus on right-censored survival problems. Detecting high-degree interactions for the estimation of survival probability is a challenging problem in survival analysis from the statistical perspective. RESULTS: We propose a new methodology, Logical Analysis of Survival Data (LASD), to identify interactions between variables (survival patterns) without any prior hypotheses. Using these set of patterns, we predict survival distributions for each observation. To evaluate LASD we select two publicly available datasets: a lung adenocarcinoma dataset (gene-expression pro.les) and the other a breast cancer dataset (clinical pro.les). The performance of LASD when compared with survival decision trees improves the cross-validation accuracy by 18% for the gene-expression dataset, and by 2% for the clinical dataset. AVAILABILITY: Executable codes will be provided upon request.