AI-Based Myocardial Segmentation and Attenuation Mapping Improved Detection of Myocardial Ischemia and Infarction on Emergency CT Angiography

Figures

• 

  Figure 1. Flowchart outlining the selection of the final study population with utilized inclusion and exclusion criteria within the defined observation window. CT = computed tomography, PE = pulmonary embolism, AAS = acute aortic syndrome, ICA = invasive coronary angiography, ECG = electrocardiography, STEMI = ST-elevation myocardial infarction, NSTEMI = non ST-elevation myocardial infarction.

• 

  Figure 2. Figure showing the heart from a CT examination without (first row; (A–C)) and with (second row; (D–F)) overlying colormaps for cases with a STEMI, NSTEMI and no myocardial infarction and vessel stenosis < 50%. Voxels with a negative deviation from the mean of the myocardium are shown in blue; voxels with a positive deviation are shown in red. The third row shows the corresponding invasive coronary angiographies (ICA) with a total proximal occlusion of the RCA for the STEMI (G), a high-grade stenosis of the LAD for the NSTEMI (H) and no hemodynamically relevant stenosis in the case with no infarction (I). Arrows indicate the location of vessel occlusion or high-grade stenosis in the angiographic images. The case with the STEMI shows a transmural hypodensity of the inferior wall (A,D). The case with the NSTEMI demonstrates a small subendocardial hypodensity (B,E). In the case with no relevant vessel stenosis, no hypodense myocardium was described (C,F).

• 

  Figure 3. (A) Percentage of readers that rated hypodense myocardium (HM) in each case without colormaps. (B) Percentage of readers that rated hypodense myocardium in each case with supportive colormaps. The colors of the bars are used only for visual distinction and do not represent additional categories.

• 

  Figure 4. (A) Percentage of readers that rated hypodense myocardium (HM) in each case for the subgroup of patients with a myocardial infarction without colormaps, and (B) with supportive colormaps. (C) Percentage of readers that rated hypodense myocardium (HM) in each case for the subgroup of patients with a STEMI without colormaps, and (D) with supportive colormaps. (E) Percentage of readers that rated hypodense myocardium (HM) in each case for the subgroup of patients with a NSTEMI without colormaps, and (F) with supportive colormaps. (G) Percentage of readers that rated hypodense myocardium (HM) in each case for the subgroup of patients without a myocardial infarction without colormaps, and (H) with supportive colormaps. The colors of the bars are used only for visual distinction and do not represent additional categories.

• 

  Figure 5. Inter-reader agreement for hypodense myocardium with on average lower kappa values without colormaps (A) and on average higher kappa values with supportive colormaps (B). On average, the use of colormaps allowed less experienced readers to achieve a higher agreement with expert readers concerning the detection of a hypodense myocardium (A,B).