Nongated Photon‐Counting Computed Tomography Angiography Detects Cardioembolic Stroke Sources and Thoracic Pathology: A Retrospective Cohort Study

Abstract

BACKGROUND

Stroke is often attributable to cardiothoracic factors, but the need for ECG gating and limited spatial resolution on conventional computed tomography often limits the detection of pathology relevant for secondary stroke prevention. The study objective was to evaluate the frequency of cardiothoracic pathology on non‐ECG‐gated photon‐counting computed tomography angiography during initial stroke imaging.

METHODS

Consecutive patients with a clinical suspicion of acute stroke, who were imaged using non‐ECG‐gated dual‐source photon‐counting computed tomography with diaphragm‐to‐vertex coverage at a comprehensive stroke center, were retrospectively included. Image quality was assessed using a 4‐point Likert scale, and images were evaluated for cardiac stroke sources and thoracic pathology. Where available, results from echocardiography were collected.

RESULTS

The study included 193 complete stroke investigations, 126 of which (65.3%; 95% CI: 58.3–71.6) had imaging‐confirmed ischemic strokes. The image quality was generally high for cardiac imaging (excellent 9.8% [95% CI: 6.4–14.9], good 50.3% [95% CI: 43.3–57.2], moderate 37.3% (95% CI: [30.8–44.3]), poor 2.6% [95% CI: 1.1–5.9]) and thoracic imaging (excellent 7.8% [95% CI: 4.8–12.4]), good 59.6% [95% CI: 52.5–66.3]), moderate 32.6% [95% CI: 26.4–39.5]), poor 0% [95% CI: 0.0–2.0]). Clinically relevant cardioembolic findings were detected in 4.7% (95% CI: 2.5–8.6) of all patients: 6 cardiac thrombi (3.1%; 95% CI: 1.4–6.6), 3 aortic valve vegetations (1.6%; 95% CI: 0.5–4.5). Other findings that could affect patient management were detected in 31.6% (95% CI: 25.5–38.5) of scans, including 14 patent foramen ovale (7.3%; 95% CI: 4.4–11.8), 13 pulmonary embolisms (6.7%; 95% CI: 4.0–11.2), 29 pleural effusions (15.0%; 95% CI: 10.7–20.7), and 8 lung malignancies (4.1%; 95% CI: 2.1–8.0). Cardiac thrombi and vegetations were exclusively found in the ischemic stroke subgroup.

CONCLUSIONS

Cardiothoracic pathology relevant to patient management or secondary stroke prevention was commonly detected with good diagnostic image quality as part of the standard stroke imaging workup with non‐ECG‐gated high‐pitch dual‐source photon‐counting computed tomography angiography.

Early identification of the etiology of ischemic stroke is crucial for secondary prevention. ¹ Attributable causes of ischemic stroke include intra‐ and extracranial atherosclerosis, cardioembolic sources, microvascular disease, and dissections. ² , ³ In approximately 25%–30% of patients with ischemic stroke, routine diagnostic investigations fail to identify an underlying etiology. ⁴ Current guidelines advocate for echocardiography to detect structural causes of cardiac embolism. ⁵ However, the timing of echocardiography may vary, and a cardioembolic focus may have dissolved at the time of imaging if delayed. ⁶ Transthoracic echocardiography (TTE) is the most frequently used method, but visualization of structural abnormalities (such as thrombi or septal defects) may be challenging. ⁷ , ⁸ Transesophageal echocardiography (TEE) exhibits better diagnostic accuracy than TTE but is less frequently performed. ⁷ , ⁸

In clinical routine care, cardiac computed tomography (CT) is always ECG gated, requiring the placement of electrodes and synchronization between the scanner and the patient's heart rhythm. ECG‐gated cardiac CT in patients with stroke has been performed and can detect high‐risk sources of cardioaortic embolism, such as intracardiac thrombi, ⁹ , ¹⁰ , ¹¹ and may suggest a superior diagnostic yield compared with TTE, ¹¹ but comes with the cost of ECG electrode placement and image acquisition time. High diagnostic accuracy using ECG‐gated cardiac CT is reported in detecting patent foramen ovale (PFO), ¹² , ¹³ although other studies appear less successful. ¹⁴ In CT, a PFO diagnosis is based on a left‐to‐right contrast shunt, differing from the typical right‐to‐left shunting seen during Valsalva in echocardiography. Most patients with PFO on TEE show a left‐to‐right shunt on ECG‐gated cardiac CT. ¹² , ¹³ Although ECG‐gated cardiac CT typically increases image acquisition time, it seems feasible to use nongated cardiac CT to detect high‐risk sources of embolism, albeit with a heightened risk of suboptimal image quality. ⁹ , ¹⁰ , ¹¹ , ¹² A recently published single‐center prospective randomized trial asserts that detection of cardioaortic thrombi is feasible at the primary stroke investigation by extending the conventional CT angiography 6 cm below the carina. ¹³

Nonstandard Abbreviations and Acronyms

EID

energy‐integrating detector

HU

Hounsfield units

LAA

left atrial appendage

PCCT

photon‐counting computed tomography

PFO

patent foramen ovale

TEE

transesophageal echocardiography

TTE

transthoracic echocardiography

CLINICAL PERSPECTIVE

What Is New?

• Non‐ECG‐gated dual‐source photon‐counting computed tomography angiography from the diaphragm to the vertex enables complementary high‐quality thoracic and cardiac imaging in the acute stroke setting.

What Are the Clinical Implications?

• In a retrospective consecutive cohort of 193 stroke investigations, clinically relevant cardioembolic findings were discovered in 4.7% (95% CI: 2.5–8.6), and other cardiothoracic findings that could affect patient management were detected in one third of scans (31.6%; 95% CI: 25.5–38.5).

• Cardiac thrombi and vegetations were identified with high diagnostic confidence, highlighting the potential of this approach as a first‐line tool for detecting cardiothoracic sources of embolism and guiding targeted secondary prevention.

Photon‐counting CT (PCCT) features a semiconductor detector that directly converts X‐ray photons into electrical signals. The detector measures individual X‐ray photons and their energies and lowers the electronic noise, making the detector performance of PCCT superior to conventional energy‐integrating detector (EID) CT. ¹⁴ PCCT thus allows for higher spatial resolution, improved contrast‐to‐noise ratio, and lower radiation dose. ¹⁵ , ¹⁶ PCCT also provides intrinsically higher sensitivity to the effects of iodine and may facilitate the management of artifacts due to beam hardening and metal. ¹⁴

Recently, PCCT has been clinically approved and implemented into clinical workflows. Given the inherent benefits of dual‐source PCCT and its rapid imaging capabilities, high‐pitch nongated arterial PCCT angiography from the diaphragm to the vertex of the skull may have a low risk of motion artifacts and provide additional clinical value in identifying cardioaortic stroke sources during initial stroke imaging, without incurring extra time or cost to the imaging procedure. We report on the diagnostic image quality and cardiothoracic imaging findings of clinical relevance in a consecutive cohort of patients with suspected stroke undergoing dual‐source PCCT as part of their diagnostic stroke workup at a regional stroke center.

METHODS

The data that support the findings of this study are available from the corresponding author upon reasonable request.

Ethical Considerations

This retrospective single‐center study was approved by the Swedish ethical review authority (approval number: 2023‐06209‐01). Informed consent was waived due to the study's retrospective nature.

Study Population

All consecutive patients with a clinical suspicion of acute stroke imaged with a clinical dual‐source PCCT system, Siemens NAEOTOM Alpha (Siemens Healthineers, Erlangen, Germany) between October 4, 2023 and April 13, 2024 at the comprehensive stroke center of Karolinska University Hospital (Solna, Sweden) were included. The majority of patients are triaged to our institution according to the prehospital Stockholm Stroke Triage System, representing a National Institutes of Health Stroke Scale score of ≥2 in the ipsilateral leg and arm (in total National Institutes of Health Stroke Scale score ≥4) upon arrival. ¹⁷ Patients were excluded if cardiothoracic coverage was not obtained during the scan.

Image Acquisition

As part of a quality improvement initiative at our institution, all patients were imaged by a dual‐source protocol tailored for rapid and comprehensive imaging, including an arterial PCCT angiography from the diaphragm to the vertex of the skull. Scans were acquired using the “RoutineFlashSpiralChildThoraxQuantum” mode (“Child” is a nominal label rather than a functional limitation) with effective tube voltage set to 90 kVp, collimation set to 144×0.40 mm, CARE Dose 4D setting turned off, effective mAs set at 250, and the shaped filter set to narrow. A pitch of 1.6 and a rotation time of 0.25 seconds were used, resulting in a CTDIvol of 6.96 mGy (32 cm). The scan direction was from feet to head, covering approximately 52 cm in an average acquisition time of 1.3 seconds, yielding an average dose length product of 346 mGy*cm (4.5 mSv). Iodixanol (Visipaque 320 mg I/mL, GE Healthcare), 80 mL, was injected intravenously at a rate of 6 mL/s, followed by a saline chaser of 80 mL NaCl (9 mg/mL) at 8 mL/s. Automatic bolus triggering was set at +100 Hounsfield units (HU) @120kVp in the descending aorta, followed by a 6‐second post‐bolus delay. The scanning parameters included an axial slice thickness of 0.4 mm, a slice increment of 0.4 mm, a 512 matrix size, a field of view reconstruction diameter of 210 mm, and images reconstructed with an Hv56 kernel using QIR 4 and Polyenergetic T3D Spectral Recon.

Imaging Analysis

The quality of the PCCT angiography was evaluated using a 4‐point Likert scale concerning motion artifacts (poor–substantial blurring, insufficient diagnostic quality; moderate–moderate blurring, moderate diagnostic quality; good–mild blurring, good diagnostic quality; excellent–no blurring, excellent diagnostic quality) in 5 distinct locations: heart chambers (overall, to detect intraluminal pathology), atrial septum (to detect atrial septal defects such as PFO), left atrial appendage (LAA) (to detect thrombi), left proximal and midsegment anterior descending artery, and lung parenchyma. The pulmonary arteries' furthest easily distinguishable branching level was noted (pulmonary trunk, lobar, segmental, sub‐segmental). The contrast opacification of the right atrium, pulmonary trunk, left atrium, left ventricle (LV), and tubular ascending aorta was trichotomized into “high” (>250 HU), “moderate” (150–250 HU), and “low” (<150 HU). ¹¹

The presence of a “certain PFO” was defined as a channel between the left and right atrium with a contrast material jet flowing between the atria. ¹⁸ A probable contrast jet or an interatrial septal aneurysm (defined as a protrusion of at least 10 mm beyond the plane of the atrial septum) with a concurrent channel‐like septum was classified as a “probable PFO.” ¹⁹ An isolated septal aneurysm or a channel‐like appearance of the interatrial septum was classified as a “possible PFO,” as both these entities are associated with PFO in about 50% of cases. ¹⁹ , ²⁰

A thrombus in the LAA, left atrium, or LV was defined as a rounded hypodense filling defect, typically measuring <100 HU. ¹² , ²¹ Valve vegetations were registered, being suggestive of endocarditis. ²¹ Other findings, such as prosthetic aortic or mitral valves, valvuloplasties, severe valvular calcifications or pannus, and suspicion of rheumatic heart disease were noted. Ascending aortic dissections and atheromatic plaque ulcerations of the ascending aorta and aortic arch proximal to the neck vessels of >2 mm were registered. ¹² , ²¹ Mixed plaques with possible ulcerations in the aorta proximal to the neck vessels were noted.

Calcifications within the proximal and mid‐segment left anterior descending artery were evaluated using a 4‐point Likert scale (severe, moderate, mild, or absent, adopted from ¹² ). The prevalence of pulmonary embolism, tumors, and pleural effusion was registered.

All scans were assessed by 2 independent readers (F.S. and A.D., a board‐certified radiologist and a board‐certified cardiac imaging specialist, each with 7 years of experience), who reached consensus decisions about image quality and/or pathology in applicable cases. An external senior cardiothoracic radiologist (R.N.P., 16 years of experience) adjudicated a small selection of cases (approximately 10%) where consensus could not be reached.

Echocardiography results were collected after PCCT angiography images had been evaluated, whenever available, including the timing and type of echocardiographic investigation performed.

Statistical Analysis

The Wilson Score interval was used to provide 95% CIs for estimates of prevalence. Statistical analyses, including measures of sensitivity and specificity, were made in RStudio (v2023.12.0, Posit Software, PBC, and R v4.3.2, The R Foundation for Statistical Computing).

RESULTS

Study Population

Of 249 consecutive stroke investigations, 193 included cardiothoracic imaging, forming the study population. Among these, 126 scans (65.3%; 95% CI: 58.3–71.6) confirmed ischemic stroke through imaging (either acute infarct visible on initial CT or diffusion weighted imaging‐positive on subsequent magnetic resonance imaging, where available), and 67 scans (34.7%; 95% CI: 28.4–41.7) showed no imaging evidence of ischemic stroke on available imaging (Figure 1, flow diagram). Accordingly, 2 groups were defined: “imaging‐confirmed stroke” (n = 126), and “others” (n = 67), the latter possibly representing a mix of nonimage‐confirmed strokes and stroke mimics. In 56 investigations, a full diagnostic stroke protocol with PCCT angiography from the diaphragm to the vertex was not performed as it was not deemed clinically motivated, thus precluding inclusion in the study. Most commonly, this was due to an intracranial hemorrhage or a patient transferred from another hospital following a CT angiography at the hospital of origin (see Figure 1). Table S1 provides details about the stroke occlusion site. The median age was 74 (interquartile range, 61–81) years, with 99 scans corresponding to male patients (51.3%; 95% CI: 44.3–58.2), and 34 (17.6%; 95% CI: 12.9–23.6) showing imaging findings consistent with prior cardiothoracic procedures (see Table 1 for population characteristics). In total, 39.4% (95% CI: 32.8–46.4) underwent subsequent TTE or TEE.

Figure 1.

Flow diagram showing included and excluded scans.

Table 1.

Population Characteristics

	Study population (n = 193)	Imaging confirmed ischemic stroke (n = 126)	Others (n = 67)
Median age, y (IQR)	74 (61–81)	75 (66–81)	72 (56–80)
Male sex, n (%)	99 (51.3) [44.3–58.2]	62 (49.2) [40.6–57.8]	37 (55.2) [43.4–66.5]
Previous cardiothoracic procedure, n (%)
No	159 (82.4) [76.4–87.1]	104 (82.5) [75.0–88.2]	55 (82.1) [71.3–89.4]
Yes	34 (17.6) [12.9–23.6]	22 (17.5) [11.8–25]	12 (17.9) [10.6–28.7]
Prosthetic valve	14 (7.3) [4.4–11.8]	6 (4.8) [2.2–10.0]	8 (11.9) [6.2–21.8]
Pacemaker	6 (3.1) [1.4–6.6]	4 (3.2) [1.2–7.9]	2 (3.0) [0.8–10.2]
Mitral valvuloplasty	5 (2.6) [1.1–5.9]	5 (4.0) [1.7–9.0]	0 (0.0) [0.0–5.4]
CABG	4 (2.1) [0.8–5.2]	3 (2.4) [0.8–6.8]	1 (1.5) [0.3–8.0]
Ascending aorta graft	3 (1.6) [0.5–4.5]	3 (2.4) [0.8–6.8]	0 (0.0) [0.0–5.4]
LVAD	1 (0.5) [0.1–2.9]	1 (0.8) [0.1–4.4]	0 (0.0) [0.0–5.4]
PTE	1 (0.5) [0.1–2.9]	0 (0.0) [0.0–3.0]	1 (1.5) [0.3–8.0]
TTE/TEE follow‐up, n (%)	76 (39.4) [32.8–46.4]	63 (50.0) [41.4–58.6]	13 (19.4) [11.7–30.4]
TTE	46 (23.8) [18.4–30.3]	35 (27.8) [20.7–36.2]	11 (16.4) [9.4–27.1]
TTE IASC	5 (2.6) [1.1–5.9]	5 (4.0) [1.7–9.0]	0 (0.0) [0.0–5.4]
TTE LVC	14 (7.3) [4.4–11.8]	12 (9.5) [5.5–15.9]	2 (3.0) [0.8–10.2]
TEE	11 (5.7) [3.2–9.9]	11 (8.7) [4.9–15.0]	0 (0.0) [0.0–5.4]
Time from PCCT to TTE/TEE, d (IQR)	2 (1–5)	2 (1–5)	1 (1–3)

Data in brackets are 95% CIs.

CABG indicates coronary artery bypass graft surgery; IASC, interatrial septum contrast agent evaluation; IQR, interquartile range; LVAD, left ventricular assist device; LVC, left ventricular contrast agent evaluation; PCCT, photon‐counting computed tomography; PTE, pulmonary thromboendarterectomy; TEE, transesophageal echocardiography; and TTE, transthoracic echocardiography.

Imaging Coverage and Contrast‐Enhancement

In 159 scans (82.4%; 95% CI: 76.4–87.1), the heart chambers were entirely/almost entirely imaged, whereas in 34 scans (17.6%; 95% CI: 12.9–23.6), the heart was not fully imaged, typically rendering the apical segments of the LV non‐assessable. The contrast intensity within the left atrium, LV and ascending aorta was generally high (>250 HU in 99.5% of scans, 95% CI: 97.1–99.9), whereas the right atrium typically had a low or moderate contrast intensity (250 HU or lower in 79.3% of scans, 95% CI: 73.0–84.4) (Table S2).

Image Quality

The overall image quality of the heart chambers was rated as excellent or good in 116 scans (60.1%; 95% CI: 53.1–66.8), moderate in 72 scans (37.3%; 95% CI: 30.8–44.3), and poor in 5 scans (2.6%; 95% CI: 1.1–5.9). Regarding motion artifacts in the LAA and atrial septum, image quality was excellent or good in 92 (47.7%; 95% CI: 40.7–54.7) and 84 (43.5%; 95% CI: 36.7–50.6) scans, respectively, moderate in 77 (39.9%; 95% CI: 33.2–46.9) and 82 (42.5%; 95% CI: 35.7–49.5) scans, and poor in 24 (12.4%; 95% CI: 8.5–17.8) and 27 (14.0%; 95% CI: 9.8–19.6) scans. The proximal and mid segments of the left anterior descending artery were considered to have excellent or good diagnostic quality in 66 scans (34.2%; 95% CI: 27.9–41.1). The lung parenchyma showed excellent or good diagnostic quality in 130 scans (67.4%; 95% CI: 60.5–73.6) and was never rated as poor quality. Pulmonary arteries were clearly distinguishable to subsegmental level in most scans (68.9%; 95% CI: 62.1–75.0) and to at least segmental level in almost all scans (97.4%; 95% CI: 94.1–98.9). Refer to Table 2 for an overview of perceived image quality. Figures 2 and 3 illustrate examples of the overall image quality of the heart and lungs, respectively.

Table 2.

Image Quality

	Study population (n = 193)	Imaging confirmed ischemic stroke (n = 126)	Others (n = 67)
Heart chambers, n (%)
Excellent	19 (9.8) [6.4–14.9]	11 (8.7) [4.9–15.0]	8 (11.9) [6.2–21.8]
Good	97 (50.3) [43.3–57.2]	66 (52.4) [43.7–60.9]	31 (46.3) [34.9–58.1]
Moderate	72 (37.3) [30.8–44.3]	46 (36.5) [28.6–45.2]	26 (38.8) [28.0–50.8]
Poor	5 (2.6) [1.1–5.9]	3 (2.4) [0.8–6.8]	2 (3.0) [0.8–10.2]
Left atrial appendage, n (%)
Excellent	26 (13.5) [9.4–19.0]	17 (13.5) [8.6–20.5]	9 (13.4) [7.2–23.6]
Good	66 (34.2) [27.9–41.1]	48 (38.1) [30.1–46.8]	18 (26.9) [17.7–38.5]
Moderate	77 (39.9) [33.2–46.9]	46 (36.5) [28.6–45.2]	31 (46.3) [34.9–58.1]
Poor	24 (12.4) [8.5–17.8]	15 (11.9) [7.3–18.7]	9 (13.4) [7.2–23.6]
Interatrial septum, n (%)
Excellent	20 (10.4) [6.8–15.5]	12 (9.5) [5.5–15.9]	8 (11.9) [6.2–21.8]
Good	64 (33.2) [26.9–40.1]	48 (38.1) [30.1–46.8]	16 (23.9) [15.3–35.3]
Moderate	82 (42.5) [35.7–49.5]	50 (39.7) [31.6–48.4]	32 (47.8) [36.3–59.5]
Poor	27 (14.0) [9.8–19.6]	16 (12.7) [8.0–19.6]	11 (16.4) [9.4–27.1]
LAD: proximal to middle segment, n (%)
Excellent	14 (7.3) [4.4–11.8]	7 (5.6) [2.7–11.0]	7 (10.4) [5.2–20.0]
Good	52 (26.9) [21.2–33.6]	42 (33.3) [25.7–41.9]	10 (14.9) [8.3–25.3]
Moderate	106 (54.9) [47.9–61.8]	66 (52.4) [43.7–60.9]	40 (59.7) [47.7–70.6]
Poor	21 (10.9) [7.2–16.1]	11 (8.7) [4.9–15.0]	10 (14.9) [8.3–25.3]
Lung parenchyma, n (%)
Excellent	15 (7.8) [4.8–12.4]	8 (6.3) [3.3–12.0]	7 (10.4) [5.2–20.0]
Good	115 (59.6) [52.5–66.3]	81 (64.3) [55.6–72.1]	34 (50.7) [39.1–62.3]
Moderate	63 (32.6) [26.4–39.5]	37 (29.4) [22.1–37.8]	26 (38.8) [28.0–50.8]
Poor	0 (0.0) [0.0–2.0]	0 (0.0) [0.0–3.0]	0 (0.0) [0.0–5.4]
Pulmonary arteries: furthest level distinguishable, n (%)
Sub‐segmental	133 (68.9) [62.1–75.0]	89 (70.6) [62.2–77.9]	44 (65.7) [53.7–75.9]
Segmental	55 (28.5) [22.6–35.2]	33 (26.2) [19.3–34.5]	22 (32.8) [22.8–44.7]
Lobar	5 (2.6) [1.1–5.9]	4 (3.2) [1.2–7.9]	1 (1.5) [0.3–8.0]

Data in brackets are 95% CIs.

LAD indicates left anterior descending artery.

Figure 2.

Axial photon‐counting computed tomography images depicting typical examples of cardiac image quality. The examples (A–D) are from male and female participants aged 63–82. A, excellent, (B) good, (C) moderate, (D): poor.

Figure 3.

Sagittal photon‐counting computed tomography images depicting typical examples of pulmonary image quality. The examples (A–C) are from male and female participants aged 69–78. A, excellent, (B) good, (C) moderate. No scan was deemed to be of poor image quality with regard to pulmonary visualization, precluding an example to be provided.

Cardioaortic Sources of Embolism

PCCT findings of potential cardioaortic sources of embolism are presented in Table 3 and elaborated on in detail in the following paragraphs. See Figure 4 for image examples of a PFO and an LAA thrombus.

Table 3.

Potential Cardioaortic Sources of Embolism

	Study population (n = 193)	Imaging confirmed ischemic stroke (n = 126)	Others (n = 67)
Patent foramen ovale, n (%)
Certain/probable	14 (7.3) [4.4–11.8]	12 (9.5) [5.5–15.9]	2 (3.0) [0.8–10.2]
Possible–aneurysm	5 (2.6) [1.1–5.9]	5 (4.0) [1.7–9.0]	0 (0.0) [0.0–5.4]
Possible–channel	44 (22.8) [17.4–29.2]	29 (23.0) [16.5–31.1]	15 (22.4) [14.1–33.7]
No	104 (53.9) [46.8–60.8]	64 (50.8) [42.2–59.4]	40 (59.7) [47.7–70.6]
N/A	26 (13.5) [9.4–19.0]	16 (12.7) [8.0–19.6]	10 (14.9) [8.3–25.3]
LAA thrombus, n (%)
Certain/probable	5 (2.6) [1.1–5.9]	5 (4.0) [1.7–9.0]	0 (0.0) [0.0–5.4]
Possible	40 (20.7) [15.6–27.0]	29 (23.0) [16.5–31.1]	11 (16.4) [9.4–27.1]
No	126 (65.3) [58.3–71.6]	79 (62.7) [54.0–70.6]	47 (70.1) [58.3–79.8]
N/A	22 (11.4) [7.6–16.7]	13 (10.3) [6.1–16.9]	9 (13.4) [7.2–23.6]
LA thrombus, n (%)
Certain/probable	0 (0.0) [0.0–2.0]	0 (0.0) [0.0–3.0]	0 (0.0) [0.0–5.4]
Possible	1 (0.5) [0.1–2.9]	1 (0.8) [0.1–4.4]	0 (0.0) [0.0–5.4]
No	185 (95.9) [92.0–97.9]	121 (96.0) [91.0–98.3]	64 (95.5) [87.6–98.5]
N/A	7 (3.6) [1.8–7.3]	4 (3.2) [1.2–7.9]	3 (4.5) [1.5–12.4]
LV thrombus, n (%)
Certain/probable	1 (0.5) [0.1–2.9]	1 (0.8) [0.1–4.4]	0 (0.0) [0.0–5.4]
Possible	2 (1.0) [0.3–3.7]	1 (0.8) [0.1–4.4]	1 (1.5) [0.3–8.0]
No	149 (77.2) [70.8–82.6]	96 (76.2) [68.0–82.8]	53 (79.1) [67.9–87.1]
N/A	41 (21.2) [16.1–27.5]	28 (22.2) [15.8–30.2]	13 (19.4) [11.7–30.4]
Other possible cardiac foci, n (%)
Yes–vegetation	3 (1.6) [0.5–4.5]	3 (2.4) [0.8–6.8] *	0 (0.0) [0.0–5.4]
Yes–other	7 (3.6) [1.8–7.3]	7 (5.6) [2.7–11.0] †	0 (0.0) [0.0–5.4]
Yes–biological valve	9 (4.7) [2.5–8.6]	3 (2.4) [0.8–6.8]	6 (9.0) [4.2–18.2]
Yes–mechanic valve	3 (1.6) [0.5–4.5]	1 (0.8) [0.1–4.4]	2 (3.0) [0.8–10.2]
Yes–severe aortic valve / mitral valve calcification	23 (11.9) [8.1–17.2]	18 (14.3) [9.2–21.5]	5 (7.5) [3.2–16.3]
No	138 (71.5) [64.8–77.4]	89 (70.6) [62.2–77.9]	49 (73.1) [61.5–82.3]
N/A	9 (4.7) [2.5–8.6]	5 (4.0) [1.7–9.0]	4 (6.0) [2.3–14.4]
Aortic source of embolism, n (%)
Yes–type A dissection	1 (0.5) [0.1–2.9]	1 (0.8) [0.1–4.4]	0 (0.0) [0.0–5.4]
Yes–aortic plaque ulceration	5 (2.6) [1.1–5.9]	5 (4.0) [1.7–9.0]	0 (0.0) [0.0–5.4]
Yes–other	2 (1.0) [0.3–3.7]	2 (1.6) [0.4–5.6] ‡	0 (0.0) [0.0–5.4]
Possible–severe aortic plaque	20 (10.4) [6.8–15.5]	19 (15.1) [9.9–22.4]	1 (1.5) [0.3–8.0]
Possible–other	2 (1.0) [0.3–3.7]	0 (0.0) [0.0–3.0]	2 (3.0) [0.8–10.2] §
No	163 (84.5) [78.7–88.9]	99 (78.6) [70.6–84.8]	64 (95.5) [87.6–98.5]

Data in brackets are 95% CIs.

^* One bio‐AVR vegetation, 1 mechanical‐AVR vegetation, 1 aortic valve vegetation

^† Five mitral valvuloplasties, 1 rheumatic mitral valve, 1 severe pannus on the aortic valve

^‡ Two ascending aorta grafts with severely irregular surface

^§ One plaque ulceration in the left subclavian artery, 1 recent ascending aorta graft without apparent luminal aberration

AVR indicates aortic valve replacement; LA, left atrium; LAA, left atrial appendage; and LV, left ventricle.

Figure 4.

Image examples. A, Photon‐counting CT, sagittal plane. The white arrow points toward a patent foramen ovale in a 73‐year‐old female. A contrast material jet is seen flowing between the atria, indicated by the white triangle. B, Photon‐counting CT, coronal plane. The white arrow points at a thrombus in the left atrial appendage in an 82‐year‐old male.

CT indicates computed tomography; LA, left atrium; LV, left ventricle; and RA, right atrium.

Cardioembolic Sources–Patent Foramen Ovale

In the total study population, 14 (7.3%; 95% CI: 4.4–11.8) certain or probable PFOs were found (1 was an iatrogenic atrial septal defect), of which 12 were in the imaging‐confirmed ischemic stroke subgroup (9.5% of the subgroup; 95% CI: 5.5–15.9). In addition, 49 possible PFOs were found (25.4%; 95% CI: 19.8–32.0), of which 34 were found in the imaging‐confirmed ischemic stroke subgroup (27.0% of the subgroup; 95% CI: 20.0–35.3).

Echocardiography was performed 17 times to determine the presence of a PFO (n = 11 TEE and n = 6 TTE). Among the 17 matched PCCT‐echocardiography pairs, 3 PCCT scans were deemed nonassessable due to poor quality. In the remaining 14 PCCT‐echocardiography pairs, echocardiography detected 4 PFOs. All of these PFOs were determined (while blinded to the echocardiography results) to be certain, probable, or possible on PCCT (Figure S1A displays an interatrial septal aneurysm with a PFO on paired contrast‐enhanced TEE). In 2 cases, PCCT revealed a channel‐like interatrial septum without a contrast jet, whereas echocardiography did not detect any PFOs (Figure S2A shows a channel‐like interatrial septum on PCCT for reference). Based on this very limited data (4 true positives, 2 false positives, 8 true negatives, 0 false negatives), PCCT demonstrated 100% sensitivity and 80% specificity in detecting PFO.

Cardioembolic Sources–Thrombi

In the subgroup of imaging‐confirmed ischemic stroke, 5 certain or probable LAA thrombi and 1 LV thrombus were found (4.8%; 95% CI: 2.2–10.0). A possible cardiac thrombus was found in 31 instances (24.6%; 95% CI: 17.9–32.8). In 29 of 31 cases, this was due to a filling defect in the LAA (See Figure S2B for an image example of a filling defect in the LAA). No certain/probable cardiac thrombi were found in the subgroup without imaging‐confirmed ischemic stroke. The LV thrombus was confirmed with TTE. None of the LAA thrombi found on PCCT were confirmed with echocardiography, as none underwent TEE. One LAA thrombus was found on TEE in a patient with a left ventricular assist device; however, in this case, the PCCT cardiac image quality was poor (LAA deemed nonassessable).

Cardioembolic Sources–Valves and Vegetations

In the imaging‐confirmed ischemic stroke subgroup, 18 cases (14.3%; 95% CI: 9.2–21.5) with severe aortic valve or mitral valve calcifications and 7 cases (5.6%; 95% CI: 2.7–11.0) with mitral valvuloplasty, severe pannus, or probable rheumatic valve disease were seen. These pathological findings were considerably more frequent compared with the group without imaging‐confirmed ischemic stroke (see Table 3).

In the subgroup of imaging‐confirmed ischemic stroke, 3 aortic valve vegetations were found (2.4%; 95% CI: 0.8–6.8), all in patients with previous cardiothoracic procedures (1 biological and 1 mechanical aortic valve replacement, 1 pacemaker patient). All 3 vegetations were confirmed with echocardiography, rendering 100% sensitivity and specificity of PCCT to detect vegetations in the imaging‐confirmed ischemic stroke group with available echocardiography (3 true positive, 0 false positive, 60 true negative, 0 false negative based on echocardiography). Figure S1B features vegetation on a biological aortic valve replacement, confirmed with TEE.

Aortic Sources of Embolism

In total, 1 type A dissection, 5 aortic plaque ulcerations, and 2 ascending aorta grafts with severe lumen irregularities were found, all within the imaging‐confirmed ischemic stroke subgroup (6.3% prevalence in this subgroup; 95% CI: 3.3–12.0). Additionally, the imaging‐confirmed ischemic stroke subgroup had a high prevalence of mixed plaques with severely irregular surfaces in the aorta proximal to the origin of the neck vessels (n = 19, 15.1% prevalence in this subgroup; 95% CI: 9.9–22.4, see Table 3).

Other Cardiothoracic Findings

In total, 150 scans showed some degree of calcification in the proximal or middle segments of the left anterior descending artery (77.7%; 95% CI: 71.3–83.0). In the imaging‐confirmed ischemic stroke subgroup, the prevalence of moderate to severe calcification was higher compared with the other subgroup (58.7%; 95% CI: 50.0– 66.9, vs 41.8%; 95% CI: 30.7–53.7). Pulmonary embolism was found in 13 scans (6.7%; 95% CI: 4.0–11.2), pleural effusion in 29 scans (15.0%; 95% CI: 10.7–20.7), and possible or probable lung tumors in 8 scans (4.1%, of which one had a known malignancy; 95% CI: 2.1–8.0). Table S3 displays a detailed breakdown of these cardiothoracic findings.

DISCUSSION

This retrospective single‐center consecutive stroke cohort study demonstrates the clinical value of nongated dual‐source PCCT angiography in identifying sources of cardioembolic ischemic stroke in routine comprehensive stroke imaging. Notably, cardiothoracic pathology was visualized with overall good‐to‐excellent image quality with clinically relevant findings for patient management and/or secondary stroke prevention in about every third patient.

Rapid identification of the underlying cause of ischemic stroke is crucial for optimal secondary stroke prevention. Current guidelines recommend prompt investigation of potential cardiac and thoracic embolic sources; however, standard diagnostic approaches often involve multiple separate investigations, which can delay definitive management. Our results suggest that it is feasible to achieve good diagnostic quality with a single modality. This integrated approach not only shortens diagnostic time, potentially allowing for earlier initiation of targeted therapy (eg, anticoagulation in the presence of intracardiac thrombi), but may also reduce the necessity for subsequent additional testing. Streamlining diagnostic workups offers practical advantages, including improved patient flow, reduced healthcare costs, and potentially better clinical outcomes through earlier and more accurate initiation of secondary preventive treatments.

Poor image quality of the heart chambers was observed in 5 scans (2.6%; 95% CI: 1.1–5.9), which is similar to the 3.3% reported in a large study on ECG‐gated EID dual‐source CT ¹¹ and better than the 9% and 12.5% reported in extensive studies on nongated EID single‐source CT. ¹⁷ , ¹⁸ In 1 scan, the poor quality was attributed to a left ventricular assist device, and in 3 scans, it was due to a high (>32) body mass index. The non‐ECG‐gated nature of the angiography in this study is also a fundamental requisite, as ECG gating takes time that could potentially delay treatment (eg, mechanical thrombectomy).

The frequency of PCCT findings consistent with possible, probable, or certain PFO in the imaging‐confirmed ischemic stroke subgroup (36.5%; 95% CI: 28.6–45.2) and in the nonconfirmed subgroup (25.4%; 95% CI: 16.5–36.9) is in line with the reported prevalence of PFO in autopsy series (weighted mean of 25%) and in patients with cryptogenic stroke (>25%). ²² Although we cautiously report high sensitivity (100%) and specificity (80%) measures in this non‐head‐to‐head comparison, our results should ideally be evaluated further in a prospective manner and in larger sample sizes, given that ECG‐gated complete cardiac cycle EID CT has shown 89.4% sensitivity and 92.3% specificity to detect PFO in patients with stroke. ²³

In the imaging‐confirmed ischemic stroke subgroup, a probable or certain cardiac thrombus was found in the LAA or LV in 4.8% (95% CI: 2.2–10.0) of scans, and a cardiac thrombus was deemed possible in an additional 24.6% (95% CI: 17.9–32.8) of scans, typically due to a filling defect in the LAA. Our results are in line with a recently published prospective randomized trial of extended EID CT angiography in patients with stroke, with confirmed LAA or LV thrombi in 5.3% of scans. ¹³ For comparison, 2 reports on cardiac EID CT to detect cardioembolic sources of ischemic stroke found a cardiac thrombus in 7.1% and 13% of patients with stroke, with the highest occurrence in the LAA. ¹² , ²¹ In both the latter reports, a 2‐phasic contrast material injection protocol was employed, where the heart was imaged at the time of a second contrast bolus. The delayed‐phase approach enables contrast material saturation within the myocardium and heart chambers, which is especially helpful for thrombus evaluation in the LAA. As multienergy imaging is an inherent benefit of PCCT, there is a prospect of exploring the role of iodine maps to increase the conspicuity of cardiac thrombi, as was already shown in the context of delayed‐phase EID CT imaging. ¹²

All 3 vegetations identified in the imaging‐confirmed ischemic stroke subgroup were subsequently confirmed with TEE. Although limited in sample size, our results are corroborated by a recent publication, manifesting cardiac CT as a complement to TEE or a valuable diagnostic alternative in patients with contraindications to TEE. ²⁴

Our results illustrate high‐quality cardiothoracic imaging in the routine clinical setting using nongated high‐pitch dual‐source PCCT angiography. 94.8% (95% CI: 90.7–97.2) of scans had high or moderate contrast opacification in the pulmonary trunk, indicating the possibility of assessing for concurrent pulmonary embolism. 6.7% (95% CI: 4.0–11.2) of scans showed evidence of pulmonary embolism, and possible/probable lung tumors were seen in 4.1% (95% CI: 2.1–8.0), similar to a previous study on cardiothoracic imaging in patients with stroke. ¹¹

One key strength of this study is that it covers a consecutive stroke cohort scanned on a dual‐source PCCT scanner with 24‐hour availability in a clinical routine setting. Our findings are relevant given the increasing availability of PCCT in routine practice. However, the monocentric retrospective design and a custom‐made PCCT angiography protocol limit generalizability. This being said, the protocol was optimized for rapid acquisition, enhanced iodine contrast visualization, and improved dose efficiency, and is made freely available for download: https://github.com/karolinskaimaging/. Apart from the additional cardiothoracic images to be assessed by the radiologist, no additional time or costs were imposed on patient management, due to the fast image acquisition (1.3 seconds from the diaphragm to the vertex) implemented into routine stroke imaging. Due to the study's retrospective nature, only a limited number of patients underwent echocardiography to evaluate for PFO and LAA thrombi, limiting the ability for multimodal head‐to‐head comparisons. Thus, a large subset of patients was not compared with a reference standard. Also, 56 out of 249 patients did not undergo cardiothoracic imaging, which may contribute to a selection bias. Contrary to previous studies with EID CT, we did not employ a 2‐phase contrast material injection protocol. The single‐phase protocol limits the diagnostic accuracy of cardiac thrombi, especially in the LAA due to slow filling, but also in the LV and left atrium due to inadequate contrast saturation within the myocardium. Although partly remedied by the rapid acquisition of high‐pitch dual‐source PCCT, small thrombi and subtle PFOs may be missed due to susceptibility to possible motion without ECG‐gating. The protocol was mainly designed to detect relevant findings in emergency stroke situations before thrombectomy.

Future prospective multicenter studies are warranted to evaluate the clinical significance of PCCT angiography, employing a 2‐phasic injection protocol, systematically incorporating TEE as a gold standard reference, and using the multienergy PCCT information to improve diagnostic utility. Moreover, ECG‐gating should be evaluated to discover the full potential of the PCCT technique and its impact on time metrics.

CONCLUSION

In conclusion, nongated high‐pitch dual‐source PCCT angiography performed from the diaphragm to the vertex of the skull typically provides a cardiothoracic investigation of high quality, at virtually no time expense. Nongated PCCT angiography is a promising technique to be used as a primary screening method for cardioembolic stroke. Moreover, the PCCT angiography provides additional cardiothoracic information of clinical relevance, such as pulmonary embolisms, malignancies, and pleural effusions.

Author Contributions

J.L. and T.G. led the study conception and design, to which all authors contributed. A.S. designed the image acquisition protocol. Material preparation and data collection were performed by F.S. and A.D., with assistance from R.N.P. Data analysis was performed by F.S. and reviewed by all authors. The first draft of the manuscript was written by F.S. and all authors commented on previous versions of the manuscript. All authors read and approved the final manuscript.

Sources of Funding

No funds, grants, or other support were received.

Disclosures

The authors have no relevant financial or nonfinancial interests to disclose.

Supporting information

Figure S1 A: An interatrial septal aneurysm is shown in a 49‐year‐old male on an axial PCCT image (left), and a PFO is depicted in the same patient on TEE (right). B: A vegetation on a biological aortic valve replacement in a 75‐year‐old female is shown on a coronal PCCT image (left). The finding was confirmed on TEE (right). LA indicates left atrium; PCCT, photon‐counting computed tomography; PFO, patent foramen ovale; RA, right atrium; and TEE, transesophageal echocardiography.

Figure S2 A (top row): A sagittal (top left) and axial (top right) PCCT image example of a “channel‐like” interatrial septum in a 73‐year‐old male. It represents an incomplete fusion of the interatrial septum, with a possible underlying PFO. B (bottom row): A sagittal (bottom left) and axial (bottom right) image example of a filling defect in the left atrial appendage in a 77‐year‐old female on PCCT. It may be caused by slow flow conditions in the LAA. LA indicates left atrium; LAA, left atrial appendage; PCCT, photon‐counting computed tomography; and PFO, patent foramen ovale.

Table S1: Occlusion site.

Table S2: Imaging characteristics.

Table S3: Cardiothoracic pathology, continued.