An open-source framework for pulmonary fissure completeness assessment

Abstract

We present an open-source framework for pulmonary fissure completeness assessment. Fissure incompleteness has been shown to associate with emphysema treatment outcomes, motivating the development of tools that facilitate completeness estimation. Generally, the task of fissure completeness assessment requires accurate detection of fissures and definition of the boundary surfaces separating the lung lobes. The framework we describe acknowledges a) the modular nature of fissure detection and lung lobe segmentation (lobe boundary detection), and b) that methods to address these challenges are varied and continually developing. It is designed to be readily deployable on existing lung lobe segmentation and fissure detection data sets. The framework consists of multiple components: a flexible quality control module that enables rapid assessment of lung lobe segmentations, an interactive lobe segmentation tool exposed through 3D Slicer for handling challenging cases, a flexible fissure representation using particles-based sampling that can handle fissure feature-strength or binary fissure detection images, and a module that performs fissure completeness estimation using voxel counting and a novel surface area estimation approach. We demonstrate the usage of the proposed framework by deploying on 100 cases exhibiting various levels of fissure completeness. We compare the two completeness level approaches and also compare to visual reads. The code is available to the community via github as part of the Chest Imaging Platform and a 3D Slicer extension module.

Keywords: Fissure completeness; Lung lobe fissures.

Figure 1.

Overlay images produced by quality control module in the Chest Imaging Platform. Top row: left lung. Bottom row: right lung.

Figure 2.

Interactive lung lobe segmentation using 3D Slicer. Left: sagittal slice in the left lung showing four manually selected fiducial points along the left oblique fissure. Right: final lung lobe segmentation based on the user-specified fiducial points (rendered in 3D in the upper-right quadrant).

Figure 3.

Diagram illustrating the approach to estimate the lobe boundary surface area intercepted by a voxel with physical coordinates in the axial plane. Here, Δ_x and Δ_y are half of the pixel spacing in the x and y directions in the axial plane. Center: axial plane projection of the triangle vertices. Right: schematic depiction of triangles in 3D. The surface area is estimated as the sum of each triangle’s area.

Figure 4.

2D schematic example of a “fissure” (solid arc from A to C) and “lobe boundary” (arc from A to E). Dashed line segments between A and B are analogous to the surface area approximating scheme described in the text.

Figure 5.

Synthetically generated image volume used in the workflow testing framework. Left and middle: sagittal slices of artificial (digitally generated) “lobes” and fissure detections (red overlay) with different levels of completeness. Right: 3D rendering of a fissure detection sagittal slice (white), particles (small blue discs), and fissure-specific mesh surfaces (right horizontal fissure surface in red and right oblique fissure surface in yellow).

Figure 6.

Fissure detection representation workflow. Left: input to particles ridge surface sampling (Gaussian-smoothed probability image derived from CNN feature enhancement). Middle: output of particles sampling operation. Right: wireframe mesh representation of surfaces derived from Delaunay triangulation of particles data.

Figure 7.

Heatmaps showing estimated surface areas for the left oblique (left), right oblique (center), and right horizontal (right) lobe boundaries for a typical case.

Figure 8.

Comparison of fissure completeness measures as assessed by particles sampling of a) estimation of surface areas (x-axis) and b) ratio of voxel counts (y-axis). In each panel, the line of identity is shown for reference.

Figure 9.

Quantitative fissure completeness estimates vs. visual scoring. Red: completeness measures using voxel counting. Black: completeness measures using surface area estimation. Visual scores are coded as: 1=absent/near absent, 2=mostly incomplete, 3=partially complete, 4=mostly complete, 5=complete/near complete.