Many chest abnormalities can lead to life-threatening diseases if not detected and monitored early. Some of these abnormalities show radiologic progression and are associated with increased mortality rates. One such group of abnormalities is interstitial lung abnormalities (ILAs). Interstitial lung abnormalities can be defined as non-dependent, parenchymal abnormalities that are often found incidentally on chest CT scans. ILAs can pose a significant risk of progressing into interstitial lung diseases (ILDs), which can severely impact the overall health of a patient.  

According to the Fleischner Society’s position paper [1], ILAs have been associated with higher mortality rates in patients with lung cancer and chronic obstructive pulmonary disease (COPD. With an increasing awareness of ILAs and their potential implications, the role of radiologists in detecting and reporting them has become more critical [2]. However, identifying these abnormalities on chest CT scans is a complex and time-consuming process, requiring careful examination of hundreds of chest CT image slices for each patient.  

To address this challenge, our deep learning approach is designed to assist physicians by automating and enhancing the detection of chest abnormalities, including ILAs. By integrating AI into the radiology workflow, our system can help radiologists identify abnormalities more efficiently. The proposed study aims to support doctors in making early, well-informed patient management decisions, ultimately leading to timely interventions and better patient outcomes.