Abstract
Early detection and appropriate management of treatment-related interstitial lung disease (ILD) are important in cancer treatment. We established an algorithm for quantifying fine crackles using machine learning and reported that the fine crackle quantitative value (FCQV) calculated by this algorithm was more sensitive than chest radiography for detecting interstitial changes. Using this algorithm, we periodically analyzed respiratory sounds in two patients with lung cancer who developed treatment-related ILDs and found that the FCQV was elevated before the diagnosis of ILD. These cases may indicate the usefulness of the FCQV in the early diagnosis of treatment-related ILDs.