NA-MIC Project WeeksThe Chest Imaging Platform (CIP) is an open-source software suite developed by the Applied Chest Imaging Laboratory (ACIL) at Brigham and Women’s Hospital (Harvard Medical School). Slicer CIP is designed for quantitative chest CT analysis, helping clinicians and researchers identify and measure phenotypes for Chronic Obstructive Pulmonary Disease (COPD), Interstitial Lung Disease (ILD), Acute Lung Injury (ALI), Pulmonary Vascular Pruning, and Lung Cancer.
To develop new modules for the Slicer Chest Imaging Platform (CIP)[1]. The first module aims to simplify CT feature extraction for skeletal rib biomarkers. The second module aims to implement SYBIl[2], a deep-learning model for lung cancer outcome predictions.
Rib Analysis Module
Input Data: Chest CT Images and TotalSegmentator [3] segmentations are required as inputs. If existing segmentations are not available, the TotalSegmentator slicer module wil be used to generate the segmentations.
Module Development: Quantitative CT metrics (Mass, Volume, and Density) will be extracted based on the definitions in the CIP Parenchyma Analysis module [1]. The feature quantifications will be modified to reflect the methodlogy performed in our research project [4]. Briefly, Mass and Volume will be extracted form all rib segmentations, then for each rib level the right and left sides will be aggregated. Using the level-aggregated metrics, Density (Mass divided by Volume) will be computed for each rib level. Finally, averaged metrics for Mass, Volume, and Density will be computed for three anatomical regions: True Rib Region (levels 2-7), False Rib Region (levels 8-10), and Floating Rib Region (levels 11-12). Rib level 1 was excluded due to inconsistent segmentations from varying FOV. Based on our results, the most stable CT Rib metrics was CT Rib Density extracted from the True Rib Region, which will be considered the main metric.
Module Outputs: The quantified CT Rib biomarkers will be displayed in a results table, a 3D reconstruction hgihlight the rib levels included in the main metric (CT Rib Density), and bar graphs to illustrate the averaged metrics (Mass and Density) from the anatomical regions.
SYBIL Module
Input Data: DICOM Chest CT Images.
Module Development: Leveraged CIP blank module template to create CIP Sybil module. Module searchers for either downloaded SYBIL repository in module directory or pip installs Sybil repository for users (will require restart on first use). Metrics follow SYBIL standard approaches and extracts the latent of the model, along with generating 6-year prediction and heatmap for the view.
Module Outputs: Extracts latent features and exports as .pkl file. Also, exports the six lung cancer predictions as a .json file. Further, generates heatmaps for the regions of interest, shown in the viewer.
Progress
Next Steps
Issues
Overview of the new Slicer CIP modules (CIP Rib Analysis and CIP SYBIL).
Demo of the CIP Rib Analysis Module.
Demo of the CIP SYBIL Module.