Cloud Hosting

The following is a summary of Development and validation of a CT-based deep learning algorithm to augment non-invasive diagnosis of idiopathic pulmonary fibrosis, published in the November 2023 issue of Pulmonology by Maddali, et al.

For a study, researchers sought to enhance the non-invasive diagnosis of idiopathic pulmonary fibrosis (IPF) by developing and validating a machine learning algorithm utilizing computed tomography (CT) scans exclusively. The aim was to improve the identification of the usual interstitial pneumonia (UIP) pattern and distinguish IPF from other interstitial lung diseases (ILD), reducing the need for invasive surgical biopsies.

A primary deep learning convolutional neural network (CNN) was employed, trained on a diverse multi-center dataset of over 2000 ILD cases with a reference standard of multidisciplinary discussion (MDD) consensus diagnosis. The algorithm was fine-tuned on a US-based multi-site cohort (n = 295) and externally validated with a separate dataset (n = 295) from European and South American sources.

In the tuning phase, the developed machine learning model demonstrated a commendable performance with an area under the receiver operating characteristic curve (AUC) of 0.87 (CI: 0.830.92) for distinguishing idiopathic pulmonary fibrosis (IPF) from other interstitial lung diseases (ILDs). The sensitivity and specificity reached 0.67 (0.570.76) and 0.90 (0.830.95). Notably, the model outperformed pre-recorded assessments conducted before multidisciplinary discussion (MDD) diagnosis, where sensitivity was only 0.31 (0.230.42), and specificity was 0.92 (0.870.95). The external test set validated the models robustness, yielding a c-statistic of 0.87 (0.830.91). Remarkably, the models performance consistency extended across diverse CT scanner manufacturers and various slice thicknesses.

The deep learning algorithm, relying solely on CT images, accurately identified IPF within ILD cases. The consistent results across diverse datasets and scanner variations suggested its potential as a valuable tool for non-invasive IPF diagnosis, offering improvements over traditional diagnostic approaches.

Source: resmedjournal.com/article/S0954-6111(23)00316-5/fulltext

See the original post:
Refining Non-Invasive Diagnosis of IPF: Development and Validation of a CT-Based Deep Learning Algorithm - Physician's Weekly