Unsupervised machine learning identifies predictive progression markers of IPF

Abstract

Objectives: To identify and evaluate predictive lung imaging markers and their pathways of change during progression of idiopathic pulmonary fibrosis (IPF) from sequential data of an IPF cohort. To test if these imaging markers predict outcome.

Methods: We studied radiological disease progression in 76 patients with IPF, including overall 190 computed tomography (CT) examinations of the chest. An algorithm identified candidates for imaging patterns marking progression by computationally clustering visual CT features. A classification algorithm selected clusters associated with radiological disease progression by testing their value for recognizing the temporal sequence of examinations. This resulted in radiological disease progression signatures, and pathways of lung tissue change accompanying progression observed across the cohort. Finally, we tested if the dynamics of marker patterns predict outcome, and performed an external validation study on a cohort from a different center.

Results: Progression marker patterns were identified and exhibited high stability in a repeatability experiment with 20 random sub-cohorts of the overall cohort. The 4 top-ranked progression markers were consistently selected as most informative for progression across all random sub-cohorts. After spatial image registration, local tracking of lung pattern transitions revealed a network of tissue transition pathways from healthy to a sequence of disease tissues. The progression markers were predictive for outcome, and the model achieved comparable results on a replication cohort.

Conclusions: Unsupervised learning can identify radiological disease progression markers that predict outcome. Local tracking of pattern transitions reveals pathways of radiological disease progression from healthy lung tissue through a sequence of diseased tissue types.

Key points: • Unsupervised learning can identify radiological disease progression markers that predict outcome in patients with idiopathic pulmonary fibrosis. • Local tracking of pattern transitions reveals pathways of radiological disease progression from healthy lung tissue through a sequence of diseased tissue types. • The progression markers achieved comparable results on a replication cohort.

Keywords: Idiopathic pulmonary fibrosis; Tomography, X-ray computed; Unsupervised machine learning.

Fig. 1

Overview of the algorithm and dataset. a First, unsupervised learning selects marker candidates, which results in most significant progression markers. b This flowchart represents the selection of enrolled scans

Fig. 2

Evaluation of the stability of the progression markers. a The pattern example among the top 4 ranked pattern. b Most informative progression markers identified by the model, and the repeatability of this ranking after 20 runs of random 95–5% patient splits. The top 4 ranked patterns are stable across all runs. The ranking of less informative patterns fluctuates across runs. c The top 4 ranked cluster volume representation from a patient at 4 different time points

Fig. 3

The survival study of Kaplan-Meier (KM) estimation of the most informative progression markers. a The KM curve based on markers of the scan B on the study cohort. b The KM curve based on markers of the scan B and the difference of the scan A and B on the study cohort. c The KM curve based on markers of the scan B on the replication cohort. d The KM curve based on markers of the scan B and the difference of the scan A and B on the replication cohort

Fig. 4

Pattern transition networks: mapping local pattern transition networks to reconstruct pathway candidates. a From the population of spatially matched follow-up pairs of lungs, we can observe local change of lung tissue from one to another pattern. (b) This enables obtaining a network of transition probabilities of lung patterns changing to others from one to the next examination time point. The matrix shows how likely a source pattern transitions to a target pattern. Red indicates high probability, blue low probability. These probabilities are generated by an underlying latent transition network that exhibits transition pathways shown in this figure. For the top ranked most informative patterns, we plot two pathways to illustrate this model. c Pathways originating from a healthy pattern (cluster 9), and (d) pathways ending in vessels and ground glass pattern (cluster 17). Arrows point at dominant directions in the graph