Skip to main page content
U.S. flag

An official website of the United States government

Dot gov

The .gov means it’s official.
Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

Https

The site is secure.
The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
[Preprint]. 2025 Jul 17:2025.06.09.25329157.
doi: 10.1101/2025.06.09.25329157.

AI-based Hepatic Steatosis Detection and Integrated Hepatic Assessment from Cardiac CT Attenuation Scans Enhances All-cause Mortality Risk Stratification: A Multi-center Study

Jirong Yi  1 , Krishna K Patel  2 , Robert J H Miller  1   3 , Anna M Marcinkiewicz  1   4 , Aakash Shanbhag  1   5 , Waseem Hijazi  1 , Naga Dharmavaram  1 , Mark Lemley  1 , Jianhang Zhou  1 , Wenhao Zhang  1 , Joanna X Liang  1 , Giselle Ramirez  1 , Valerie Builoff  1 , Leandro Slipczuk  6 , Mark Travin  6 , Erick Alexanderson  7 , Isabel Carvajal-Juarez  7 , René R S Packard  8 , Mouaz Al-Mallah  9 , Terrence D Ruddy  10 , Andrew J Einstein  11 , Attila Feher  12 , Edward J Miller  12 , Wanda Acampa  13 , Stacey Knight  14   15 , Viet T Le  14   15 , Steve Mason  14   15 , Vinicius F Calsavara  1 , Panithaya Chareonthaitawee  16 , Samuel Wopperer  16 , Alan Kwan  1 , Lixia Wang  1 , Daniel S Berman  1 , Damini Dey  1 , Marcelo Di Carli  17 , Piotr J Slomka  1
Affiliations

AI-based Hepatic Steatosis Detection and Integrated Hepatic Assessment from Cardiac CT Attenuation Scans Enhances All-cause Mortality Risk Stratification: A Multi-center Study

Jirong Yi et al. medRxiv. .

Abstract

Background: Hepatic steatosis (HS) is a common cardiometabolic risk factor frequently present but under-diagnosed in patients with suspected or known coronary artery disease. We used artificial intelligence (AI) to automatically quantify hepatic tissue measures for identifying HS from CT attenuation correction (CTAC) scans during myocardial perfusion imaging (MPI) and evaluate their added prognostic value for all-cause mortality prediction.

Methods: This study included 27039 consecutive patients [57% male] with MPI scans from nine sites. We used an AI model to segment liver and spleen on low dose CTAC scans and quantify the liver measures, and the difference of liver minus spleen (LmS) measures. HS was defined as mean liver attenuation < 40 Hounsfield units (HU) or LmS attenuation < -10 HU. Additionally, we used seven sites to develop an AI liver risk index (LIRI) for comprehensive hepatic assessment by integrating the hepatic measures and two external sites to validate its improved prognostic value and generalizability for all-cause mortality prediction over HS.

Findings: Median (interquartile range [IQR]) age was 67 [58, 75] years and body mass index (BMI) was 29.5 [25.5, 34.7] kg/m2, with diabetes in 8950 (33%) patients. The algorithm identified HS in 6579 (24%) patients. During median [IQR] follow-up of 3.58 [1.86, 5.15] years, 4836 (18%) patients died. HS was associated with increased mortality risk overall (adjusted hazard ratio (HR): 1.14 [1.05, 1.24], p=0.0016) and in subpopulations. LIRI provided higher prognostic value than HS after adjustments overall (adjusted HR 1.5 [1.32, 1.69], p<0.0001 vs HR 1.16 [1.02, 1.31], p=0.0204) and in subpopulations.

Interpretations: AI-based hepatic measures automatically identify HS from CTAC scans in patients undergoing MPI without additional radiation dose or physician interaction. Integrated liver assessment combining multiple hepatic imaging measures improved risk stratification for all-cause mortality.

PubMed Disclaimer

Conflict of interest statement

KKP reports an institutional research grant from Jubilant DraxImage and research support from American College of Cardiology Geriatric Cardiology council. RJM has received grant funding and consulting fees from Pfizer as well as grant funding from Alberta Innovates. AMM has received consulting fees from APQ Health. TDR has received research grant support from GE Healthcare and Advanced Accelerator Applications. AJE has received a speaker’s fee from Ionetix, consulting fees from W. L. Gore & Associates and Artrya, authorship fees from Wolters Kluwer Healthcare—UpToDate, and served on scientific advisory boards for Axcellant and Canon Medical Systems USA; his institution has grants/grants pending from Alexion, Attralus, BridgeBio, Canon Medical Systems USA, GE HealthCare, Intellia Therapeutics, Ionis Pharmaceuticals, Neovasc, Pfizer, Roche Medical Systems, Synektik, and W. L. Gore & Associates. EJM has received grant support and serves as a consultant for GE Healthcare. DD, DB, and PS have equity interest in APQ Health. Inc. DB and PS participate in software royalties for QPS software at Cedars-Sinai Medical Center. DB also served as a consultant for GE Healthcare. PS has received research grant support from Siemens Medical Systems and consulting fees from Synektik S.A. The remaining authors have declared no competing interests.

Figures

Figure 1.
Figure 1.. Overview of study design.
A previously validated model (TotalSegmentator) was used to segment liver and spleen for hepatic quantification and hepatic steatosis identification which was subsequently used for risk stratification. Additionally, an eXtreme Gradient Boosting (XGBoost) model was developed and validated for integrated liver assessment, achieving further improvement over individual hepatic measure in risk stratification. * Adjusted for clinical variables, MPI metrics, and all hepatic measures. # Adjusted for clinical and MPI variables. MPI – myocardial perfusion imaging, EHR – electronic health record, CTAC – computed tomography attenuation correction.
Figure 2.
Figure 2.. Study cohort creation flowchart with 27039 patients from four SPECT/CT sites and five PET/CT sites.
SPECT – single-photon emission computed tomography, CT – computed tomography, PET – positron emission tomography, CTAC – computed tomography for attenuation correction
Figure 3.
Figure 3.. Relation between hazard ratio and hepatic measure.
A1, B1, C1: adjusted hazard ratios at different liver minus spleen (LmS) attenuation values with population mean as the reference. A2, B2, C2: adjusted hazard ratios at different liver attenuation values with population mean as the reference. A1-A2: plots from entire population. B1-B2: plots from female subpopulation; C1-C2: plots from male subpopulation. Factors for adjustment included: age, sex, body mass index, diabetes mellitus, dyslipidemia, family history of coronary artery disease, hypotension, smoking, log10(CAC score+1), left ventricular ejection fraction, stress TPD, liver attenuation (or LmS attenuation), liver coefficient of variation (CoV), LmS CoV, LmS entropy, and LmS standard deviation. HU – Hounsfield unit, LmS – liver minus spleen, TPD – total perfusion deficit.
Figure 4.
Figure 4.. Adjusted Kaplan-Meier curves of hepatic measures for all-cause mortality prognosis.
A-D: Kaplan-Meier curves of HSIC-based hepatic steatosis (HS), liver coefficient of variation (CoV), LmS (liver minus spleen) entropy, and LmS CoV for all-cause mortality risk stratification, respectively in MPI cohort of 27039 patients. *Adjusted for clinical/perfusion factors (sex, age, body mass index, hypertension, diabetes mellitus, dyslipidemia, family history of coronary artery disease, smoking, stress total perfusion deficit, left ventricle ejection fraction, coronary artery calcium score, and imaging modality) and all the other hepatic quantifications. Low liver CoV/low LmS entropy/low LmS SD was reference group. Patients without hepatic steatosis measures were in reference group. BMI – body mass index, CAD – coronary artery disease, TPD – total perfusion deficit, LVEF – left ventricular ejection fraction, CAC – coronary artery calcium, LmS – liver minus spleen, MPI – myocardial perfusion imaging, HSIC – hepatic steatosis imaging criterion
Figure 5.
Figure 5.. Examples of liver and spleen segmentation from computed tomography attenuation correction (CTAC) scans.
A1, B1: axial view slice, and segmentation of liver and spleen from a CTAC scan of a male in his 60s with body mass index 23.4 kg/m2, liver attenuation −6 Hounsfield unit (HU), spleen attenuation 25 HU, and difference of liver and spleen (LmS) attenuation −31 HU. The hepatic measure implied the patient had hepatic steatosis (defined as liver attenuation < 40 HU or difference of liver minus spleen attenuation < −10 HU). A2, B2: axial view slice, and segmentation of liver and spleen from a CTAC scan of a male in his 60s with body mass index 21.4 kg/m2, liver attenuation 52 Hounsfield unit (HU), spleen attenuation 39 HU, and LmS attenuation 13 HU. BMI – body mass index.
Figure 6.
Figure 6.. Kaplan-Meier curves of hepatic steatosis imaging criterion (HSIC)-based hepatic steatosis and liver risk index (LIRI) for risk stratification of all-cause mortality in external sites.
HS was defined as liver attenuation < 40 HU or difference of liver minus spleen attenuation < −10 HU) in the external sites. # Adjusted for clinical and perfusion variables. Liver risk index (LIRI) was produced by an eXtreme Gradient Boosting model.
See this image and copyright information in PMC

Similar articles

See all similar articles

References

    1. Al-Mallah MH, 2022 Bateman TM, Branch KR, et al. ASNC/AAPM/SCCT/SNMMI guideline for the use of CT in hybrid nuclear/CT cardiac imaging. J Nucl Cardiol 2022; 29: 3491–3535. 2022/September/03. DOI: 10.1007/s12350-022-03089-z. - DOI - PubMed
    1. Gowd BM, Heller GV and Parker MW. Stress-only SPECT myocardial perfusion imaging: a review. J Nucl Cardiol 2014; 21: 1200–1212. 2014/July/10. DOI: 10.1007/s12350-014-9944-y. - DOI - PubMed
    1. Al-Mallah M, Alwan M, Al Rifai M, et al. Cardiac positron emission tomography and other modalities for coronary artery disease assessment: A snapshot from the medicare data. J Nucl Cardiol 2024; 41: 102030. 2024/September/05. DOI: 10.1016/j.nuclcard.2024.102030. - DOI - PubMed
    1. McCollough CH, Bushberg JT, Fletcher JG, et al. Answers to Common Questions About the Use and Safety of CT Scans. Mayo Clin Proc 2015; 90: 1380–1392. 2015/October/06. DOI: 10.1016/j.mayocp.2015.07.011. - DOI - PubMed
    1. Pakdaman MN, Rozanski A and Berman DS. Incidental coronary calcifications on routine chest CT: Clinical implications. Trends Cardiovas Med 2017; 27: 475–480. DOI: 10.1016/j.tcm.2017.04.004. - DOI - PubMed

Publication types

LinkOut - more resources