Assessing Lung Fibrosis with ML-Assisted Minimally Invasive OCT Imaging

This paper presents a combined optical coherence tomography (OCT) imaging/machine learning (ML) technique for real-time analysis of lung tissue morphology to determine the presence and level of invasiveness of idiopathic lung fibrosis (ILF). This is an important clinical problem as misdiagnosis is common, resulting in patient exposure to costly and invasive procedures and substantial use of healthcare resources. Therefore, biopsy is needed to confirm or rule out radiological findings. Videoscopic-assisted thoracoscopic wedge biopsy (VATS) under general anesthesia is typically necessary to obtain enough tissue to make an accurate diagnosis. This kind of biopsy involves the placement of several tubes through the chest wall, one of which is used to cut off a piece of lung to send for evaluation. The removed tissue is examined histopathologically by microscopy to confirm the presence and the pattern of fibrosis. However, VATS pulmonary biopsy can have multiple side effects, including inflammation, tissue morbidity, and severe bleeding, which further degrade the quality of life for the patient. Furthermore, the results are not immediately available, requiring tissue processing and analysis. Here, we report an initial attempt of using ML-assisted polarization sensitive OCT (PS-OCT) imaging for lung fibrosis assessment. This approach has been preliminarily tested on a rat model of lung fibrosis. Our preliminary results show that ML-assisted PS-OCT imaging can detect the presence of ILF with an average of 77% accuracy and 89% specificity.