The lungs are made up of lobes either completely or partially separated by fissures. Regions where the the lobes are fully isolated are called complete fissures, whereas partially separated regions are called incomplete fissures. The ratio of complete to total fissure area is known as fissure integrity. Fissure integrity has been shown to be a biomarker for collateral ventilation and may impact disease progression and lung biomechanics. 

Our team began our work with a standard image analysis technique for fissure integrity analysis (ISBI 2020 Paper). After its promise, our team developed a deep learning method to automatically compute fissure integrity from computed tomography images. This method works by combining density features with fissure class probability features derived from FissureNet (a fissure/lobe segmentation network).

The output of the network is an integrity image denoting regions of completeness and incompleteness across the fissure surface. This image can be used to directly compute a numerical integrity score or to locate regions of incompleteness for procedure screening.

IntegrityNet: a deep learning approach for pulmonary fissure integrity classification

For more information about fissure integrity, or how to join this project, please contact Zac Althof