Researchers have developed a comprehensive, combined deep learning and machine learning algorithm that can be used to easily, rapidly, and noninvasively diagnose and differentiate idiopathic pulmonary fibrosis (IPF) from various other interstitial lung diseases (ILDs), as published in Respirology.

The researchers used noninvasive clinical examination and high-resolution computed tomography (HRCT) to develop a multimodal artificial intelligence (AI) tool to diagnose IPF by integrating analysis of clinical information, examination, and computed tomography images, without input from a multidisciplinary discussion or invasive surgical lung biopsies.

One-thousand and sixty-eight patients with chronic ILD were evaluated between April 2007 and March 2017 at an ILD referral center in Japan. Eligible patients were evaluated using the usual noninvasive examinations for ILD: pulmonary function tests, bronchoalveolar lavage, and serologic tests.

Continue Reading

Read more about IPF experimental therapies

Chest HRCT images were obtained for all patients. On HRCT images, areas of honeycombing and/or reticular pattern with peripheral traction bronchiectasis were classified as suggestive of IPF. Areas suggestive of non-IPF were predominant consolidation, extensive pure ground-glass opacity, extensive mosaic attenuation, and/or diffuse nodules or cysts.

Statistical analysis of the study showed that IPF diagnosed by the machine learning algorithm had a prognostic discriminatory ability equivalent to that of multidisciplinary discussion-IPF and surgical lung biopsies.

This algorithm, in combination with multidisciplinary discussion, can aid with the early diagnosis and timely management of patients with IPF to preserve lung function. While the algorithm can be used to identify, and appropriately initiate timely antifibrotic drugs to manage these patients, it may also allow some patients to avoid surgical lung biopsy, thereby reducing acute exacerbation and death.

The high accuracy, sensitivity, and specificity of this algorithm is a new milestone in the diagnosis and management of IPF.


Furukawa T, Oyama S, Yokota H, et al. A comprehensible machine learning tool to differentially diagnose idiopathic pulmonary fibrosis from other chronic interstitial lung diseasesRespirology. Published online June 13, 2022. doi:10.1111/resp.14310