Impact of slice thickness on CACS calculation with virtual non-contrast in photon-counting CT

Abstract

Background

This study aims to investigate the feasibility of coronary artery calcium scoring (CACS) calculating from PureCalcium virtual non-iodine algorithm on photon-counting detector CT (PCD-CT) and the potential impact of different section thickness, level of virtual monoenergetic images (VMIs), and quantum iterative reconstruction (QIR) on the accuracy of CACS quantification.

Materials and methods

A total of 123 patients who underwent coronary CT angiography on PCD-CT with a separate true non-contrast CACS (CACS_TNC) scan were prospectively included. Agatston scores were calculated from the PureCalcium algorithm (CACS_PC) using a section thickness of 3 mm–1.5 mm, different VMI (55–75 kilo-electron volt (keV)) and QIR (strength 1,4) levels, respectively. CACS_TNC at 70 keV and QIR 2 were used as reference standards. Differences in CACS of different reconstructions section thicknesses, various keV levels, and QIR strength were compared using the Wilcoxon rank sum test with Bonferroni correction. The intraclass correlation coefficients (ICCs) and Bland-Altman analysis were conducted to assessed the agreement. The agreement of plaque burden groups (based on CACS) at different reconstruction parameters was evaluated using weighted Cohen kappa.

Results

At all investigated section thickness, VMI, and QIR levels, the CACS_PC were strongly correlated with CACS_TNC (ICC: 0.94–0.98, P < 0.001 for all). There were no statistical differences in CACS between CACS_PC at 3 mm section thickness, 60/65 keV (QIR1/4), and at 1.5 mm section thickness with 55 keV (QIR1/4), compared with CACS_TNC. The smallest CACS bias was observed at a 1.5 mm section thickness, 55 keV, QIR 1, with mean bias of 2.4; LoA (IQR: −182.7, 187.4). CACS_PC correctly identified 105 of 123 participants (85.4%) into the corresponding plaque burden group using CACS_TNC as the referent standard (excellent agreement, κ = 0.904).

Conclusion

CACS derived from the PureCalcium algorithm with optimized reconstruction parameters shows excellent correlation with true non-contrast scans derived values. Thus, it is may possible to use the PureCalcium virtual non-iodine algorithm to replace the true non-contrast scans for CACS quantification, without additional radiation dose exposure.

Introduction

Coronary artery calcium score (CACS), which is obtained from non-contrast CT scan, has been demonstrated as an independent prognostic predictor for all-cause mortality in asymptomatic individuals [1]. The progression of coronary calcification remained as an independent predictor for all-cause mortality after adjusting for baseline CACS, interval time of scan, and clinical risk factors [2]. Nowadays, CACS is recommended at a class IIa level of the 2019 AHA guidelines for patients with borderline/intermediate risk and is helpful in guiding clinical management and treatment [3, 4].

In routine clinical examinations, a separate non-contrast CT scan before coronary CT angiography (CCTA), was conducted to quantify the CACS, which may increase the examination time and lead to extra radiation dose. With the advent of dual-energy and spectral imaging techniques, several studies [5–7] have explored the assessment of the CACS on virtual non-contrast (VNC) images. These studies using dual-energy techniques, and have demonstrated high correlations but underestimated CACS compared to the true non-contrast (TNC) images. With the advent of photon-counting detector CT (PCD-CT), it has been shown to have higher spatial resolution and reduced radiation dose. Additionally, PCD-CT data contain differentiated spectral information, which can be used for multi-material decomposition [8].

In this context, a novel virtual noniodine (VNI) algorithm has been developed to create PureCalcium reconstructions on PCD-CT data. Previous studies have demonstrated that the PureCalcium algorithm achieves significantly higher accuracy in CACS quantification than VNC reconstructions [9]. Of note, current virtual non-iodine reconstructions consistently demonstrate a systematic underestimation of coronary artery calcium scores compared to true non-contrast reference standards. While previous investigations have examined the individual and combined effects of virtual monoenergetic image energy levels and iterative reconstruction strength, these studies have critically overlooked the fundamental parameter of slice thickness [10, 11]. This omission represents a substantial limitation in the existing literature, given that slice thickness directly governs partial volume effects and significantly impacts the detectability of small calcifications. Consequently, the optimal tripartite combination of slice thickness, keV level, and iterative reconstruction strength for minimizing discrepancies between virtual and true non-contrast calcium scoring remains undetermined. The systematic investigation of these three interdependent parameters is therefore essential for establishing standardized reconstruction protocols that can enable reliable cardiovascular risk assessment and facilitate the clinical adoption of virtual non-contrast calcium scoring.In this study, we aimed to assess the impact of different VMIs, QIR levels, and section thickness on the accuracy of CACS quantification on the PureCalcium algorithm on PCD-CT.

Materials and methods

Study patients

A total of 123 patients who underwent separate TNC and contrast-enhanced photon-counting CCTA due to suspected coronary artery disease between January 2024 and April 2024 were prospectively enrolled. Inclusion criteria included subjects older than 18 who underwent CCTA examinations by standard spatial resolution scanning. Exclusion criteria were as follows: (1) patients with CACS = 0; (2) patients who had a history of coronary revascularization (percutaneous coronary intervention or coronary artery bypass graft). This study was approved by our ethics committee (YJ-2024-044-1), and written informed consent was obtained.

Data acquisition

All patients were scanned using a PCD-CT (Naeotom Alpha, software version VA50, Siemens Healthineers). TNC scans were performed using the prospective ECG-gated sequential technique, the tube voltage was set at 120 kVp, and the tube current-time product was set to an IQ level of 19. CCTA scans were selected as prospective and retrospective based on the participants’ heart rate and rhythm. The tube voltage was set at 120 kVp, and the tube current-time product was set to an IQ level of 64. The contrast materials were applied using bolus tracking with an enhancement threshold of 100 HU in the descending aorta and a time delay of 6 s. A total of 45 to 60 mL of an iodinated contrast agent (350 mg iohexol/mL, Omnipaque; GE HealthCare) was injected at a flow rate of 4.5 mL/sec. Subsequently, 40 mL of saline was injected with the same rate. If not clinically contraindicated, 0.4 mg of nitroglycerin was sublingually administered approximately 5 min before the scan [12].

Image reconstruction

TNC images were reconstructed using a section thickness of 3.0 mm with an increment of 1.5 mm at 70 keV, QIR 2, and a recommended kernel Qr36, as defined by the vendor’s clinical calcium scoring protocol. The CCTA images were post-processed using a novel VNI algorithm (PureCalcium, Siemens Healthineers) with Qr36 kernel. For all of the patient data, VMIs were reconstructed at 5-keV intervals from 55 to 75 keV with QIR level 1 or 4 for each keV level and section thickness of 3.0 mm with an increment of 1.5 mm or section thickness of 1.5 mm with an increment of 1.0 mm.

Calcium scoring analysis

The TNC and PureCalcium datasets were evaluated on a semi-automated workflow (CT CaScoring, syngo.via VB60; Siemens Healthineers). Calcified lesion attenuation exceeding 130 HU and a minimum area of 0.5 mm² was automatically identified as calcium. Manual adjustments were made as necessary. The calcium scoring calculation was performed by two radiologists (H.X.Z. and Y.E.Z., with > 10 years of experience in cardiovascular imaging) on anonymized data sets provided on a syngo.via server. They were blinded to clinical information and independently assessed all data. Disagreements in calcium scoring between the two readers were resolved by consensus reading involving a third senior radiologist (with > 15 years of experience). A direct comparison was performed between the PureCalcium algorithm (CACS_PC) using prior reconstruction parameters (55 keV/QIR 1 and 60 keV/QIR 4, 3.0 mm section thickness) [9] and true non-contrast CT (CACS_TNC, 70 keV/QIR 2) to determine the patients with (CACS_TNC> 0) that was not detected by the previous reconstruction protocol. The smallest mean difference of CACS determined the combination of optimal reconstruction parameters for calcium detection compared to the true non-contrast scan.

The burden of calcium plaque in each patient was classified according to the Coronary Artery Disease Reporting System (P0 = score of 0 [none]; P1 = 1–100 [mild]; P2 = 101–300 [moderate]; P3 = 301–999 [severe]; P4 > = 1000 [extensive]) [10].

Statistical analysis

Continuous data were described as mean ± SD, while categorical data were expressed as frequencies and proportions. The Kolmogorov-Smirnov test was used to test the continuous data for normality. Spearman’s correlation coefficient (r) was used to assess the correlation between CACS_TNC and CACS_PC. The differences in calculated CACS with different reconstruction methods were compared with the referent standard (CACS_TNC) using the Friedman test with Bonferroni correction. Intraclass correlation coefficient (ICC) and Bland-Altman analysis were used for agreement assessment. The agreement of plaque burden classification between CACS_TNC and CACS_PC was analyzed using Cohen’s kappa (κ) statistics. Statistical analysis was performed using SPSS 27 (IBM Corporation, Armonk, NY) and MedCalc 22.0 (MedCalc, Ostend, Belgium).

Results

Patient characteristics

A total of 123 patients, including 78 males, were enrolled in this study. The mean age was 69.9 ± 9.5 years, with a range of 45–89 years. Table 1 summarizes the demographic characteristics of the included patients in detailed.

Table 1.

Characteristics of the enrolled patients

Characteristic	Value
Male/Female	78(63)/45(37)
Age, years	69.9 ± 9.5 (45–89)
Heart rate during CCTA, bpm	68.0 ± 17.3 (58–123)
Body mass index, kg/m2	25.1 ± 2.4
Total CACS	109.7 (36.95, 485.95)
Risk factors, n (%)
Hypertension	96 (78)
Hyperlipidemia	77 (63)
Diabetes mellitus	41 (33)
Smoking	76 (62)
Family history of coronary artery disease	32 (42)

Values are mean ± SD, median (interquartile range), or n (%). CCTA, coronary CT angiography; CACS, coronary artery calcium scoring

Comparison of CACS_TNC and CACS_PC using different reconstruction parameters

The median value of CACS_TNC was 109.7 (IQR: 36.9–496.9). With the reconstruction parameters of 3 mm section thickness and an increment of 1.5 mm, the value of median CACS_PC ranged from 83.2 (IQR: 5.5–399.7; 75 keV, QIR 4) to 132.5 (IQR: 20.2–570.1; 55 keV, QIR 1) depending on different level of VMI and QIR. For the reconstruction parameters of 1.5 mm section thickness and an increment of 1 mm, the median value of CACS_PC ranged from 76.9 (IQR: 10.8–364.5; 75 keV, QIR 4) to 107.4 (IQR: 17.0–474.6; 55 kV, QIR 1). Regardless of whether the reconstruction section thickness is 3 mm–1.5 mm, CACS_PC strongly correlates with CACS_TNC (r = 0.93 and r = 0.96, respectively, P < 0.001) and shows excellent agreement (ICC between 0.97 and 0.98 for all) for different keV and QIR levels. No statistical differences were observed in CACS_PC at 3 mm section thickness, 55 keV (QIR4), 60/65 keV (QIR1/4), and at 1.5 mm section thickness with 55 keV (QIR1/4), 60 keV (QIR1) compared with CACS_TNC (Fig. 1). Detailed results of correlation and agreement analysis are illustrated in Table 2. The smallest mean bias was obtained at 1.5 mm section thickness, 55 keV with QIR 1 (CACS: 107.5 [IQR: 18.9–480.4] compared to CACS_TNC: 109.7 [IQR: 36.9–496.9] mean bias, 2.3; LoA, (− 182.7/187.4)) (Fig. 2). Regardless of section thickness and QIR level, CACS_PC values were gradually decreased with increased keV values (P < 0.001 for all) (Fig. 1).

Fig. 1.

Impact of different section thickness, VMI and QIR levels compared with CACS_TNC. Both CACS_TNC and CACSpc are illustrated as median. ns indicates not significant. ***P < 0.001; *P < 0.05; ns > = 0.05. QIR, quantum iterative reconstruction; VMI, virtual monoenergetic image; CACS, coronary artery calcium scoring

Table 2.

Comparison of CACS_TNC and CACS_PC with different section thicknesses, VMI, and QIR

Reconstruction methods	Section thickness /increment(mm/mm)	VMI (keV)	QIR	Agatston Score (IQR)	Friedman P	r	ICC	Bias	LoA
TNC	3/1.5	70	2	109.7 (36.87,496.92)	-	-	-	-	-
Pure Calcium	3/1.5	55	1	132.5 (20.25,570.15)	0.002	0.96	0.97	48.63	-206.86/304.15
			4	126.4 (18.87,552.82)	>0.99	0.95	0.98	43.77	-197.32/284.87
		60	1	121.6 (17.12,513.02)	>0.99	0.96	0.98	19.89	-185.98/225.77
			4	112.5 (11.37,502.67)	>0.99	0.95	0.98	15.21	-180.13/210.57
		65	1	108.4 (13.47,473.87)	>0.99	0.95	0.98	-5.56	-187.44/176.31
			4	94.1 (7.47,470.07)	0.025	0.93	0.98	-13.78	-203.21/175.65
		70	1	100.2 (11.72,438.65)	<0.001	0.95	0.98	-28.07	-214.63/158.48
			4	89.1 (5.50,438.20)	<0.001	0.94	0.98	-34.86	-228.95/159.23
		75	1	89.7 (9.65, 408.85)	<0.001	0.95	0.98	-47.60	-256.27/161.07
			4	83.2 (5.52,399.70)	<0.001	0.94	0.98	-54.33	-271.89/163.23
	1.5/1.0	55	1	107.5 (18.90,480.37)	>0.99	0.96	0.98	2.35	-182.68/187.39
			4	107.4 (17.02, 474.55)	>0.99	0.95	0.98	-7.31	-186.34/171.71
		60	1	99.8 (17.12,452.80)	>0.99	0.96	0.98	-18.45	-190.94/154.04
			4	99.4 (15.90,438.70)	0.001	0.95	0.98	-27.86	-213.66/157.93
		65	1	93.1 (15.05,429.60)	0.001	0.96	0.98	-36.89	-226.26/152.48
			4	90.1 (14.72,412.90)	0.001	0.95	0.98	-46.17	-248.44/156.10
		70	1	89.3 (63.16,441.24)	<0.001	0.96	0.98	-52.38	-264.78/160.00
			4	80.7 (13.55,381.72)	<0.001	0.95	0.98	-62.29	-287.98/163.39
		75	1	86.3 (11.92,389.10)	<0.001	0.95	0.98	-64.67	-297.70/168.34
			4	76.9(10.77, 364.47)	<0.001	0.96	0.98	-75.24	-326.30/175.82

Values are median (interquartile range)

TNC indicates true noncontrast; VMI, virtual monoenergetic images; QIR, quantum iterative reconstruction; ICC, intraclass correlation; LoA, limits of agreement

P value < 0.05 was considered statistically significant

Fig. 2.

Bland-Altman analysis of CACS compared between TNC and PureCalicum algorithm at 1.5 mm,55 keV, QIR1. QIR, quantum iterative reconstruction; TNC, true noncontrast; CACS, coronary artery calcium scoring

Agreement between the 2 readers was excellent for both CACS_PC (ICC = 0.99 for all) and CACS_TNC (ICC = 0.98 for all).

Comparison of standard purecalcium and optimized purecalcium algorithms for tiny calcium detection

Of the 123 participants included, 21 (17.7%) participants had CACS_TNC >0 but were not detected on standard PureCalcium parameters reconstructions. The median CACS_TNC was 12.4 (IQR: 6.4–19.35). Of those, 13 (61.9%) participants were detected using optimized PureCalcium reconstructions with a section thickness of 1.5 mm and an increment of 1.0 mm, 55 keV, and QIR 1. Figure 3 shows representative cases of CACS_TNC and section thickness 3.0 mm standard CACS_PC compared with section thickness 1.5 mm CACS_PC using the optimized reconstruction.

Fig. 3.

Example cases of axial CT reconstructions of CACS. Comparing CACS_TNC (left column: 70 keV, quantum iterative reconstruction QIR 2), CACS_PC at standard reconstruction (middle column: 3 mm/1.5 mm,55 keV, QIR 1 and 60 keV, QIR 4, as previously recommended) and CACS_PC at the optimized reconstruction (right column: 1.5 mm/1 mm,55 keV, QIR 1). The upper row (patient 1) shows CT scans in a 75-year-old male and the lower row (patient 2) in a 61-year-old male. The optimized reconstruction improved the detection of tiny calcified plaques (circles) initially detected in TNC scans but not in standard PureCalcium reconstructions. LCX, left Circumflex Artery; RCA, right coronary artery; CACS, coronary artery calcium scoring; QIR, quantum iterative reconstruction; TNC, true noncontrast

Impact of differences in CACS on plaque burden classification

In the P classification of the Coronary Artery Disease Reporting System, 56, 26, 32, and 9 patients were classified as P1, P2, P3, and P4, respectively, using CACS_TNC as the reference standard. No statistical difference was observed in P classification overall between CACS_TNC and CACS_PC images (P = 0.06). Of note, 18 patients had different P classification results between CACS_PC and CACS_TNC. On PureCalcium images, 8 patients were misdiagnosed as P0 instead of P1 on TNC images. In these patients, the median CACS_PC was 3.95 (range: 0.3 to 10.2). Additionally, 3 patients were misclassified as P1 instead of P2, and 1 was misclassified as P3 instead of P2. P2 and P4 were misdiagnosed instead of P3 in 1 patient and 4 patients, respectively. Furthermore, 1 patient was misclassified as P3 instead of P4 (Tables 3 and 4) (Fig. 4). The categorical agreement of plaque classification between CACS_TNC and CACS_PC was excellent (κ = 0.87).

Table 3.

CACS distribution in TNC and purecalcium reconstruction

Scan Type and CACS
CACS burden	PureCalcium images	TNC images
0	8 (6)	0
1-100	51 (41)	56 (46)
101–300	23 (19)	26 (21)
301–999	29 (24)	32 (26)
> 1000	12 (10)	9 (7)

Data are numbers of patients, with percentages in parentheses. TNC, true noncontrast; CACS, coronary artery calcium scoring

Table 4.

CAD-RADS P classification based on CACS_TNC images or CACS_PC images

CAD-RADS P Classification	TNC P1	TNC P2	TNC P3	TNC P4	Total
PureCalcium P0	8	0	0	0	8
PureCalcium P1	48	3	0	0	51
PureCalcium P2	0	22	1	0	23
PureCalcium P3	0	1	27	1	29
PureCalcium P4	0	0	4	8	12
Total	56	26	32	9	123

CAD-RADS, Coronary Artery Disease Reporting and Data System; CACS, coronary artery calcium scoring

Fig. 4.

Sankey diagrams of change in Plaque Burden Classification at TNC and PureCalcium. A total of 8(6%) patients were misdiagnosed as P0 instead of P1 on Purecalcium images. 3 (2%)patients were misclassified as P1 instead of P2, and 1(0.8%) patient were misclassified as P3 instead of P2. P2 and P4 were misdiagnosed instead of P3 in 1(0.8%) and 4 (3%)participants, respectively. 1(0.8%) participant was misclassified as P3 instead of P4.TNC, true noncontrast

Discussion

Our study aimed to evaluate the feasibility of CACS quantification based on the PureCalcium algorithm from contrast-enhanced CCTA and the effects of different section thickness, VMI, and QIR levels during imaging reconstruction on the accuracy of CACS calculation. The findings of our study are as follows: First, the value of CACS_PC showed excellent correlation and agreement compared to CASC_TNC. Second, the value of CACS_PC at a reconstruction parameter of a section thickness of 1.5 mm, 55 kV (QIR 1), is closest to CACS_TNC.

In this study, only QIR1 and QIR4 were used for imaging reconstruction [7], consistent with previous studies that used different imaging techniques to evaluate the accuracy and consistency of calcium imaging. Compared with the energy integral detector CT, PCD-CT has higher spatial resolution and can improve dose efficiency. PCD-CT can utilize higher spectral separation technology for multi-substance separation and generate advanced spectral post-processed images, such as pure calcium images, which can significantly improve the accuracy of CACS calculation both in vitro and in vivo studies [10, 12– 13]. However, standard PureCalcium reconstruction is unable to detect very low-density calcifications accurately. This may be attributed to the following reasons: the limited detectability of low-density calcifications and partial volume effects due to the section thickness of 3 mm [14–17]. Vliegenthart et al. showed that the calcium volumes were higher in 1.5 mm slices than in 3 mm in vitro and in vivo [18]. Georg et al. observed that a thinner section was associated with increased CACS values [19]. Our study demonstrated that the 1.5 mm slice thickness can detect more tiny calcification, which can guide the use of this reconstruction parameter to detect small and low-density plaques.

A key finding regarding the thickness-VMIs that minimises the bias in CACS has been identified. For 3-mm slice thickness, 65 keV resulted in the lowest bias; while 1.5-mm thickness were optimised at 55 keV. This elucidates the unique physical trade-offs in the virtual non-contrast technique for CACS. For 3-mm slices, partial volume effects is a dominant limiting factor. Small or low-attenuation calcifications are susceptible to being averaged with surrounding soft tissue, reducing their apparent CT attenuation [20]. A higher monoenergetic level (65 keV) provides crucial benefits: better suppression of residual intravascular iodine and an improved signal-to-noise ratio. These effects enhance the specificity of calcium identification by mitigating false-positive classifications arising from misregistration or noise. Conversely, 1.5-mm slices have better spatial resolution, allowing for more accurate representation of small calcifications. Lowering the monoenergetic energy to 55 keV significantly elevates the contrast-to-noise ratio, optimising the sensitivity for detecting subtle and low-density calcifications.

According to CACS_PC, the agreement on risk categorization was acceptable. In this study, 85% of patients’ CACS_PC burdens were correctly categorized using CACS_TNC as the reference. It is essential for risk stratification and management recommendations. Although most CACS_PC misdiagnoses occur at P1, which may result in false-negative CACS, our study partially addressed this issue by employing 1.5 mm sections. So, CACS calculation on the PureCalcium algorithm with a reconstruction section of 1.5 mm may be a more accurate way for CACS quantification.

The observed discrepancies in individual patient CACS grading—encompassing both over- and under-estimation—are inherent to the spectral material decomposition process.The predominance of overestimation can be attributed to incomplete spectral separation in regions exhibiting high iodine concentrations. In such cases, residual contrast signal may be erroneously classified as calcification. Concurrently, elevated image noise at lower virtual monoenergetic levels and edge-blurring effects from high-strength iterative reconstruction have been shown to artificially amplify calcium measurements. Conversely, underestimation occurs as a result of inadvertent signal subtraction of low-attenuation calcifications whose spectral profiles overlap with iodine-enhanced soft tissue, particularly at lower energy levels. Furthermore, the derived CACS_PC imaging are susceptible to signal depression from decomposition inaccuracies or image noise. This can cause the calculated attenuation of truly calcified voxels to fall below the 130-HU threshold, thereby leading to underestimation.

These technical trade-offs directly inform differentiated clinical protocols.When the primary clinical objective is the detection of early or minimal coronary calcification—such as in high-risk primary prevention screening—the 1.5/1 mm and 55-keV reconstruction protocol is recommended.

Although this study has demonstrated the excellent correlation between CACS_PC and CACS_TNC and showed improved identification of tiny coronary calcifications by using 1.5 mm section thickness images, several limitations should be noted. Firstly, there were no detailed analytical results based on different QIR levels; only QIR1 and QIR4 were used in this study. The CACS calculated from other levels of QIRs should be further investigated. Secondly, the quantification of CACS can be affected by various factors; only VMIs, iterative reconstruction levels, and slice parameters were investigated; other parameters, such as different densities of calcification in terms of position, breath-hold depth, and heart rate, require further investigation. Thirdly, a relatively small number of participants were included for analysis; the consistency between CACS_PC and CACS_TNC in larger-scale populations with different severities of calcification should be further investigated. Fourthly, this study is a single center and only involved one CT vendor; the results should be interpreted carefully in other CT vendors with different virtual noniodine algorithms.

In conclusion, the CACS calculated from the PureCalcium algorithm showed excellent correlations with true non-contrast images. The slice thickness, VMI, and QIR levels can affect the results of CACS. Compared to CACS_TNC, the lower CACS_PC can be mitigated by using thinner sections and lower keV reconstructions. This method facilitates accurate PureCalcium-based assessment of CACS in clinical practice and potentially eliminates the need for separate non-enhanced scans, thereby reducing radiation dose.

Abbreviations

CACS

Coronary artery calcium scoring

PCD-CT

Photon-counting detector CT

VMIs

Virtual monoenergetic images

QIR

Quantum iterative reconstruction

VNC

Virtual non-contrast

TNC

True non-contrast

CCTA

Coronary CT angiography

ICC

Intraclass correlation coefficient

Author contributions

Qiuju hu: Writing original draft, Conceptualization, Project administration, Writing - review & editing. Huixin zhang: Data curation, Formal analysis, Bangju guo: Conceptualization, review & editing. Dongsheng Jin: Resource, Meirong Sun: Data curation JiliangJilang Chen: MethodologySong Luo: Investigation, Resources, Software, Writing - review & editing. Yane Zhao: Formal analysis, Methodology, Writing - review & editingGuang-ming Lu: Investigation, Resource.

Funding

This research received no external funding.

Data availability

The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.

Declarations

Ethical approval

This study was approved by the Ethics Committee of Geriatric Hospital of Nanjing Medical University(YJ-2024-044-1) in accordance with the Declaration of Helsinki.

Consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.