An LSTM-based approach for predicting idiopathic pulmonary fibrosis progression

Pulmonary fibrosis is a progressive lung disease that occurs when lung tissues get scarred and damaged. Although this condition cannot be completely treated, early identification and prediction of its progression can assist to keep it under control. Since this disease can occur without any cause it is termed “Idiopathic”. This disease can cause shortness of breath, fatigue, a dry cough, etc., and lead to death if left uncared. The objective of this paper is to use the patient’s HRCT images from the CT scanner, forced vital capacity (FVC) assessed with a spirometer, and other patient information like sex, smoking status, and so on to predict the severity of idiopathic pulmonary fibrosis progression in the lungs. Nowadays, Machine Learning plays a significant part in the healthcare sector for predicting and diagnosing various diseases, image segmentation, drug discovery, etc. The LSTM (Long Short Term Memory) model is utilized in this work to predict disease progression. The LSTM is a kind of RNN (Recurrent neural network) that is effectively used for predicting time series data and for sequence prediction problems. This model predicts the future values of FVC measurements through which we can know the patient’s severity of the decline.