Quantitative Detection of Pericardial Adhesions Using Four-Dimensional Computed Tomography: A Novel Motion-Based Analysis Framework

Figures

• 

  Figure 1. Overview of the cardiac motion analysis pipeline for EAT and pericardium. The heatmap utilizes color gradients to visualize the motion characteristics of the cardiac surface. As illustrated in the color spectrum of ‘3, motion disparity’, red regions correspond to areas of minimal motion amplitude, whereas blue regions indicate zones of pronounced, substantial movement. t* represents the cardiac phase at which the cumulative motion difference between the pericardium and EAT reaches its peak value.

• 

  Figure 2. EAT segmentation workflow. (a) Coronary CTA image showing pericardium as a thin white line. (b) Manual segmentation of pericardial region (green). (c) Threshold-based segmentation of adipose tissue within pericardial mask (yellow). (d) Reconstructed 3D surfaces of EAT.

• 

  Figure 3. Distribution of motion disparity across cardiac phases in ROI surrounding the LAD. (a) Visualization of EAT–pericardium motion differences. The motion difference magnitude is color-coded on the EAT surface, with cooler colors indicating larger displacement differences between the corresponding EAT and pericardial points, representing no pericardial adhesion. The greatest relative motion difference occurs during early systole(indicated by the red dashed box). When regions appear closer to red, it indicates weak relative motion, suggesting possible pericardial adhesions. (b) KS statistics’ matrix of nine cardiac phases. (c) Histograms show the distribution of the motion difference’s magnitude at each cardiac phase. The red curve represents the phase with maximum motion disparity, which corresponds to the early systolic phase of the heart (indicated by the red dashed box area in the ECG). The blue curves show other phases. X-axis: normalized displacement differences (range 0.0–1.0); Y-axis: frequency count.

• 

  Figure 4. Visualization of EAT–pericardium motion differences in patients with pericardial adhesion. (a) The spatial distribution of EAT–pericardium motion differences was visualized across the entire epicardial fat region. The phase with maximum motion disparity (marked by a red dashed box) was selected based on the normalized displacement differences, with a magnified display focusing on the ROI surrounding the LAD for detailed examination. (b) This is an intraoperative photograph from a preoperative case (patient No. 1). The patient’s previous surgery, performed one year prior, consisted of a Morrow procedure, MVR, and CABG with the left internal mammary artery to the LAD. During the current operation, extensive pericardial adhesions were noted in the surgical field, particularly around the anastomotic site.

• 

  Figure 5. Frequency distribution analysis of adhesion and non-adhesion patterns under varying motion differences. These distribution curves present the relationship between the motion and area metrics, where the X-axis denotes the relative motion difference as a normalized velocity indicator, and the Y-axis represents the frequency distribution quantified by the cardiac surface voxel counts. The Y-axis frequency indicator is expressed as a normalized area metric, with each point along the curve reflecting the normalized area distribution corresponding to its specific velocity magnitude. (a) The frequency distribution of adhesion patterns for seven experimental cases (Case1–Case7) shows exponential decay with increasing motion difference and peak frequencies near zero motion. (b) Frequency distribution of non-adhesion patterns for 13 experimental cases (Case8–Case20), demonstrating distinct peaks at motion differences between 0.5 and 0.6, with Case15 exhibiting the highest frequency (~5000). Both analyses reveal contrasting behavioral patterns in adhesion and non-adhesion characteristics across different motion ranges (0–1.0).