Joyce H. Keyak
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Joyce H. Keyak has a total of 113 co-authors. Here are the top co-authors by number of shared publications:
Harry B Skinner
Bone-Rad Therapeutics, Inc., Corona del Mar, CA
5 shared publications- Thomas F. Lang
Department of Radiology, University of California, San Francisco, CA 94143, USA
4 shared publications - Weihua Zhou
Michigan Technological University
4 shared publications Stephen A Rossi
3 shared publicationsTadashi S Kaneko
From the *Department of Orthopaedic Surgery, the †Department of Biomedical Engineering, and the ‡Department of Mechanical and Aerospace Engineering, University of California, Irvine, CA; and the §Chao Family Comprehensive Cancer Center, University of California Irvine Medical Center, Orange, CA
3 shared publicationsA. Sigmarsdottir
Icelandic Heart Association Research Institute, Kopavogur, Iceland
2 shared publications
Latest Publications
- 8 July 2026
- Published by IOP Publishing in Biomedical Physics & Engineering Express
- Vol. 12 (4) , 045013
Feasibility of opportunistic peripheral bone mineral density and structural quantification using an ultra-low dose multi-detector CT
Quantitative computed tomography (QCT) enables accurate assessment of volumetric bone mineral density (vBMD) and bone structure. Peripheral QCT (pQCT) and high-resolution pQCT (HR-pQCT) have been proposed as lower-dose alternatives to conventional QCT; however, their limited availability restricts widespread clinical and research use. To investigate the feasibility of using ultra–low-dose multi-detector computed tomography (MDCT) with a hybrid iterative reconstruction (HIR) algorithm for quantifying tibial vBMD, using synchronous liquid and solid calibration phantoms. Fifty volunteers (28 women and 22 men; age range: 30–60 yr; T -score ⩾ −1) without metabolic bone disorders, were scanned using a Philips-Brilliance MDCT scanner. To minimize radiation exposure, vBMD was quantified at the 38% distal tibial site using either a liquid calibration phantom (50, 100, and 200 mg cm −3 dipotassium hydrogen phosphate [K2HPO4]) or a solid calibration phantom (50, 100, and 250 mg cm −3 calcium hydroxyapatite [CHA]). A cross-calibration equation was applied to derive equivalent vBMD values between phantom types. Cortical bone structure was evaluated using a semi-automated segmentation approach. Following cross-calibration, BMDK2HPO4 = 1.34 BMDCHA + 13.43, mean cortical vBMD values were 1188.07 ± 29.38 mg cm −3 (CHA) and 1600.92 ± 39.51 mg cm −3 (K2HPO4). The CHA-based cortical vBMD values were consistent with solid-based pCT measurements reported in the literature. Mean cortical thickness was approximately 5 mm. The estimated effective radiation dose is approximately 10 µ Sv, corresponding to daily background radiation levels. This preliminary study demonstrates the feasibility of using HIR-based ultra–low-dose MDCT for peripheral tibial vBMD and structural assessment at the 38% distal site. This approach offers a promising opportunistic alternative in clinical and research settings where pQCT or HR-pQCT systems are unavailable.
- 8 July 2026
- Published by IOP Publishing in Biomedical Physics & Engineering Express
- Vol. 12 (4) , 045013
Feasibility of opportunistic peripheral bone mineral density and structural quantification using an ultra-low dose multi-detector CT
Quantitative computed tomography (QCT) enables accurate assessment of volumetric bone mineral density (vBMD) and bone structure. Peripheral QCT (pQCT) and high-resolution pQCT (HR-pQCT) have been proposed as lower-dose alternatives to conventional QCT; however, their limited availability restricts widespread clinical and research use. To investigate the feasibility of using ultra–low-dose multi-detector computed tomography (MDCT) with a hybrid iterative reconstruction (HIR) algorithm for quantifying tibial vBMD, using synchronous liquid and solid calibration phantoms. Fifty volunteers (28 women and 22 men; age range: 30–60 yr; T -score ⩾ −1) without metabolic bone disorders, were scanned using a Philips-Brilliance MDCT scanner. To minimize radiation exposure, vBMD was quantified at the 38% distal tibial site using either a liquid calibration phantom (50, 100, and 200 mg cm −3 dipotassium hydrogen phosphate [K2HPO4]) or a solid calibration phantom (50, 100, and 250 mg cm −3 calcium hydroxyapatite [CHA]). A cross-calibration equation was applied to derive equivalent vBMD values between phantom types. Cortical bone structure was evaluated using a semi-automated segmentation approach. Following cross-calibration, BMDK2HPO4 = 1.34 BMDCHA + 13.43, mean cortical vBMD values were 1188.07 ± 29.38 mg cm −3 (CHA) and 1600.92 ± 39.51 mg cm −3 (K2HPO4). The CHA-based cortical vBMD values were consistent with solid-based pCT measurements reported in the literature. Mean cortical thickness was approximately 5 mm. The estimated effective radiation dose is approximately 10 µ Sv, corresponding to daily background radiation levels. This preliminary study demonstrates the feasibility of using HIR-based ultra–low-dose MDCT for peripheral tibial vBMD and structural assessment at the 38% distal site. This approach offers a promising opportunistic alternative in clinical and research settings where pQCT or HR-pQCT systems are unavailable.
- 30 April 2026
- Published by Elsevier in Biomedical Signal Processing and Control
- Vol. 122
PF-DAformer: proximal femur segmentation via domain adaptive transformer for dual-center QCT
- 30 April 2026
- Published by Elsevier in Biomedical Signal Processing and Control
- Vol. 122
PF-DAformer: proximal femur segmentation via domain adaptive transformer for dual-center QCT
- 19 September 2025
- Published by Springer Nature in Journal of Imaging Informatics in Medicine
- Vol. 39 (3) , 2290-2308
A New Method of Modeling the Multi-stage Decision-Making Process of CRT Using Machine Learning with Uncertainty Quantification
Current machine learning-based (ML) models usually attempt to utilize all available patient data to predict patient outcomes while ignoring the associated cost and time for data acquisition. The purpose of this study is to create a multi-stage ML model to predict cardiac resynchronization therapy (CRT) response for heart failure (HF) patients. This model exploits uncertainty quantification to recommend additional collection of single-photon emission computed tomography myocardial perfusion imaging (SPECT MPI) variables if baseline clinical variables and features from electrocardiogram (ECG) are not sufficient. Two hundred eighteen patients who underwent rest-gated SPECT MPI were enrolled in this study. CRT response was defined as an increase in left ventricular ejection fraction (LVEF) > 5% at a 6 ± 1 month follow-up. A multi-stage ML model was created by combining two ensemble models: Ensemble 1 was trained with clinical variables and ECG; Ensemble 2 included Ensemble 1 plus SPECT MPI features. Uncertainty quantification from Ensemble 1 allowed for multi-stage decision-making to determine if the acquisition of SPECT data for a patient is necessary. The performance of the multi-stage model was compared with that of Ensemble models 1 and 2. The response rate for CRT was 55.5% (n = 121) with overall male gender 61.0% (n = 133), an average age of 62.0 ± 11.8, and LVEF of 27.7 ± 11.0. The multi-stage model performed similarly to Ensemble 2 (which utilized the additional SPECT data) with AUC of 0.75 vs. 0.77, accuracy of 0.71 vs. 0.69, sensitivity of 0.70 vs. 0.72, and specificity 0.72 vs. 0.65, respectively. However, the multi-stage model only required SPECT MPI data for 52.7% of the patients across all folds. By using rule-based logic stemming from uncertainty quantification, the multi-stage model was able to reduce the need for additional SPECT MPI data acquisition without significantly sacrificing performance.
- 19 September 2025
- Published by Springer Nature in Journal of Imaging Informatics in Medicine
- Vol. 39 (3) , 2290-2308
A New Method of Modeling the Multi-stage Decision-Making Process of CRT Using Machine Learning with Uncertainty Quantification
Current machine learning-based (ML) models usually attempt to utilize all available patient data to predict patient outcomes while ignoring the associated cost and time for data acquisition. The purpose of this study is to create a multi-stage ML model to predict cardiac resynchronization therapy (CRT) response for heart failure (HF) patients. This model exploits uncertainty quantification to recommend additional collection of single-photon emission computed tomography myocardial perfusion imaging (SPECT MPI) variables if baseline clinical variables and features from electrocardiogram (ECG) are not sufficient. Two hundred eighteen patients who underwent rest-gated SPECT MPI were enrolled in this study. CRT response was defined as an increase in left ventricular ejection fraction (LVEF) > 5% at a 6 ± 1 month follow-up. A multi-stage ML model was created by combining two ensemble models: Ensemble 1 was trained with clinical variables and ECG; Ensemble 2 included Ensemble 1 plus SPECT MPI features. Uncertainty quantification from Ensemble 1 allowed for multi-stage decision-making to determine if the acquisition of SPECT data for a patient is necessary. The performance of the multi-stage model was compared with that of Ensemble models 1 and 2. The response rate for CRT was 55.5% (n = 121) with overall male gender 61.0% (n = 133), an average age of 62.0 ± 11.8, and LVEF of 27.7 ± 11.0. The multi-stage model performed similarly to Ensemble 2 (which utilized the additional SPECT data) with AUC of 0.75 vs. 0.77, accuracy of 0.71 vs. 0.69, sensitivity of 0.70 vs. 0.72, and specificity 0.72 vs. 0.65, respectively. However, the multi-stage model only required SPECT MPI data for 52.7% of the patients across all folds. By using rule-based logic stemming from uncertainty quantification, the multi-stage model was able to reduce the need for additional SPECT MPI data acquisition without significantly sacrificing performance.
- Open Access
A new method of modeling the multi-stage decision-making process of CRT using machine learning with uncertainty quantification
Aims. The purpose of this study is to create a multi-stage machine learning model to predict cardiac resynchronization therapy (CRT) response for heart failure (HF) patients. This model exploits uncertainty quantification to recommend additional collection of single-photon emission computed tomography myocardial perfusion imaging (SPECT MPI) variables if baseline clinical variables and features from electrocardiogram (ECG) are not sufficient. Methods. 218 patients who underwent rest-gated SPECT MPI were enrolled in this study. CRT response was defined as an increase in left ventricular ejection fraction (LVEF) > 5% at a 6+-1 month follow-up. A multi-stage ML model was created by combining two ensemble models: Ensemble 1 was trained with clinical variables and ECG; Ensemble 2 included Ensemble 1 plus SPECT MPI features. Uncertainty quantification from Ensemble 1 allowed for multi-stage decision-making to determine if the acquisition of SPECT data for a patient is necessary. The performance of the multi-stage model was compared with that of Ensemble models 1 and 2. Results. The response rate for CRT was 55.5% (n = 121) with overall male gender 61.0% (n = 133), an average age of 62.0+-11.8, and LVEF of 27.7+-11.0. The multi-stage model performed similarly to Ensemble 2 (which utilized the additional SPECT data) with AUC of 0.75 vs. 0.77, accuracy of 0.71 vs. 0.69, sensitivity of 0.70 vs. 0.72, and specificity 0.72 vs. 0.65, respectively. However, the multi-stage model only required SPECT MPI data for 52.7% of the patients across all folds. Conclusions. By using rule-based logic stemming from uncertainty quantification, the multi-stage model was able to reduce the need for additional SPECT MPI data acquisition without sacrificing performance.
Open AccessA new method of modeling the multi-stage decision-making process of CRT using machine learning with uncertainty quantification
Aims. The purpose of this study is to create a multi-stage machine learning model to predict cardiac resynchronization therapy (CRT) response for heart failure (HF) patients. This model exploits uncertainty quantification to recommend additional collection of single-photon emission computed tomography myocardial perfusion imaging (SPECT MPI) variables if baseline clinical variables and features from electrocardiogram (ECG) are not sufficient. Methods. 218 patients who underwent rest-gated SPECT MPI were enrolled in this study. CRT response was defined as an increase in left ventricular ejection fraction (LVEF) > 5% at a 6+-1 month follow-up. A multi-stage ML model was created by combining two ensemble models: Ensemble 1 was trained with clinical variables and ECG; Ensemble 2 included Ensemble 1 plus SPECT MPI features. Uncertainty quantification from Ensemble 1 allowed for multi-stage decision-making to determine if the acquisition of SPECT data for a patient is necessary. The performance of the multi-stage model was compared with that of Ensemble models 1 and 2. Results. The response rate for CRT was 55.5% (n = 121) with overall male gender 61.0% (n = 133), an average age of 62.0+-11.8, and LVEF of 27.7+-11.0. The multi-stage model performed similarly to Ensemble 2 (which utilized the additional SPECT data) with AUC of 0.75 vs. 0.77, accuracy of 0.71 vs. 0.69, sensitivity of 0.70 vs. 0.72, and specificity 0.72 vs. 0.65, respectively. However, the multi-stage model only required SPECT MPI data for 52.7% of the patients across all folds. Conclusions. By using rule-based logic stemming from uncertainty quantification, the multi-stage model was able to reduce the need for additional SPECT MPI data acquisition without sacrificing performance.
- 1 April 2022
- Published by IOP Publishing in Physics in Medicine & Biology
- Vol. 67 (7) , 075007
A preliminary safety assessment of vertebral augmentation with 32 P brachytherapy bone cement
Comprehensive treatment for vertebral metastatic lesions commonly involves vertebral augmentation (vertebroplasty or kyphoplasty) to relieve pain and stabilize the spine followed by multiple sessions of radiotherapy. We propose to combine vertebral augmentation and radiotherapy into a single treatment by adding 32 P, a β -emitting radionuclide, to bone cement, thereby enabling spinal brachytherapy to be performed without irradiating the spinal cord. The goal of this study was to address key dosimetry and safety questions prior to performing extensive animal studies. The 32 P was in the form of hydroxyapatite powder activated by neutron bombardment in a nuclear reactor. We performed ex vivo dosimetry experiments to establish criteria for safe placement of the cement within the sheep vertebral body. In an in vivo study, we treated three control ewes and three experimental ewes with brachytherapy cement containing 2.23–3.03 mCi 32 P ml −1 to identify the preferred surgical approach, to determine if 32 P leaches from the cement and into the blood, urine, or feces, and to identify unexpected adverse effects. Our ex vivo experiments showed that cement with 4 mCi 32 P ml −1 could be safely implanted in the vertebral body if the cement surface is at least 4 mm from the spinal cord in sheep and 5 mm from the spinal cord in humans. In vivo , a lateral retroperitoneal surgical approach, ventral to the transverse processes, was identified as easy to perform while allowing a safe distance to the spinal cord. The blood, urine, and feces of the sheep did not contain detectable levels of 32 P, and the sheep did not experience any neurologic or other adverse effects from the brachytherapy cement. These results demonstrate, on a preliminary level, the relative safety of this brachytherapy cement and support additional development and testing.
- 1 April 2022
- Published by IOP Publishing in Physics in Medicine & Biology
- Vol. 67 (7) , 075007
A preliminary safety assessment of vertebral augmentation with 32 P brachytherapy bone cement
Comprehensive treatment for vertebral metastatic lesions commonly involves vertebral augmentation (vertebroplasty or kyphoplasty) to relieve pain and stabilize the spine followed by multiple sessions of radiotherapy. We propose to combine vertebral augmentation and radiotherapy into a single treatment by adding 32 P, a β -emitting radionuclide, to bone cement, thereby enabling spinal brachytherapy to be performed without irradiating the spinal cord. The goal of this study was to address key dosimetry and safety questions prior to performing extensive animal studies. The 32 P was in the form of hydroxyapatite powder activated by neutron bombardment in a nuclear reactor. We performed ex vivo dosimetry experiments to establish criteria for safe placement of the cement within the sheep vertebral body. In an in vivo study, we treated three control ewes and three experimental ewes with brachytherapy cement containing 2.23–3.03 mCi 32 P ml −1 to identify the preferred surgical approach, to determine if 32 P leaches from the cement and into the blood, urine, or feces, and to identify unexpected adverse effects. Our ex vivo experiments showed that cement with 4 mCi 32 P ml −1 could be safely implanted in the vertebral body if the cement surface is at least 4 mm from the spinal cord in sheep and 5 mm from the spinal cord in humans. In vivo , a lateral retroperitoneal surgical approach, ventral to the transverse processes, was identified as easy to perform while allowing a safe distance to the spinal cord. The blood, urine, and feces of the sheep did not contain detectable levels of 32 P, and the sheep did not experience any neurologic or other adverse effects from the brachytherapy cement. These results demonstrate, on a preliminary level, the relative safety of this brachytherapy cement and support additional development and testing.
- Open Access
- 1 January 2022
- Published by Elsevier in SSRN Electronic Journal
A New Hip Fracture Risk Index Derived from FEA-Computed Proximal Femur Fracture Loads and Energies-to-Failure
Hip fracture risk assessment is an important but challenging task. Quantitative CT-based patient-specific finite element (FE) analysis (FEA) incorporates bone geometry and the three-dimensional distribution of bone density in the proximal femur to compute the force (fracture load) and energy necessary to break the proximal femur in a particular loading condition. Although the fracture loads and energies-to-failure for different loading conditions are individually associated with incident hip fracture, and are mutually correlated, they each provide different structural information about the proximal femur that can influence a subject’s overall fracture risk. To obtain a more robust measure of fracture risk, we used principal component analysis (PCA) to develop a global FEA-computed fracture risk index that incorporates the FEA-computed yield and ultimate failure loads and energies-to-failure in four loading conditions (single-limb stance and impact from a fall onto the posterior, posterolateral, and lateral aspects of the greater trochanter) of 110 hip fracture subjects and 235 age- and sex-matched control subjects from the AGES-Reykjavik study. We found that the first principal component (PC1) of the FE parameters was the only significant predictor of hip fracture (p-value < 0.001); we refer to PC1 as the global FEA-computed fracture risk index. To evaluate this fracture risk index, we considered the areal bone mineral density (aBMD) and six covariates including age, sex, height, weight, health status, and bone medication status. Using a logistic regression model, we determined if prediction performance for hip fracture using PC1 differed from that using FE parameters combined by stratified random resampling with respect to hip fracture status. We also compared the performance of predicting hip fracture by the logistic regression model including PC1 with the FRAX model. The results showed that the average of the area under the receiver operating characteristic curve (AUC) using PC1 (0.776) was always higher than that using all FE parameters combined (0.737) in the male subjects (p-value < 0.001). The AUC of PC1 and AUC of the FE parameters combined were not significantly different in the female subjects (p-value = 0.211) or in all subjects (p-value = 0.159). The AUC values using PC1 (0.754 in all subjects, 0.825 in males and 0.71 in females) were greater than the respective values using FRAX (0.651 in the whole sample, 0.705 in males and 0.623 in females) with p-value < 0.01. Therefore, the global FEA-computed fracture risk index based on PCA includes information about hip fracture incidence beyond that of FRAX.
- 1 January 2022
- Published by Elsevier in SSRN Electronic Journal
Open AccessA New Hip Fracture Risk Index Derived from FEA-Computed Proximal Femur Fracture Loads and Energies-to-Failure
Hip fracture risk assessment is an important but challenging task. Quantitative CT-based patient-specific finite element (FE) analysis (FEA) incorporates bone geometry and the three-dimensional distribution of bone density in the proximal femur to compute the force (fracture load) and energy necessary to break the proximal femur in a particular loading condition. Although the fracture loads and energies-to-failure for different loading conditions are individually associated with incident hip fracture, and are mutually correlated, they each provide different structural information about the proximal femur that can influence a subject’s overall fracture risk. To obtain a more robust measure of fracture risk, we used principal component analysis (PCA) to develop a global FEA-computed fracture risk index that incorporates the FEA-computed yield and ultimate failure loads and energies-to-failure in four loading conditions (single-limb stance and impact from a fall onto the posterior, posterolateral, and lateral aspects of the greater trochanter) of 110 hip fracture subjects and 235 age- and sex-matched control subjects from the AGES-Reykjavik study. We found that the first principal component (PC1) of the FE parameters was the only significant predictor of hip fracture (p-value < 0.001); we refer to PC1 as the global FEA-computed fracture risk index. To evaluate this fracture risk index, we considered the areal bone mineral density (aBMD) and six covariates including age, sex, height, weight, health status, and bone medication status. Using a logistic regression model, we determined if prediction performance for hip fracture using PC1 differed from that using FE parameters combined by stratified random resampling with respect to hip fracture status. We also compared the performance of predicting hip fracture by the logistic regression model including PC1 with the FRAX model. The results showed that the average of the area under the receiver operating characteristic curve (AUC) using PC1 (0.776) was always higher than that using all FE parameters combined (0.737) in the male subjects (p-value < 0.001). The AUC of PC1 and AUC of the FE parameters combined were not significantly different in the female subjects (p-value = 0.211) or in all subjects (p-value = 0.159). The AUC values using PC1 (0.754 in all subjects, 0.825 in males and 0.71 in females) were greater than the respective values using FRAX (0.651 in the whole sample, 0.705 in males and 0.623 in females) with p-value < 0.01. Therefore, the global FEA-computed fracture risk index based on PCA includes information about hip fracture incidence beyond that of FRAX.
- 1 February 2020
- Published by Elsevier in Journal of Biomechanics
- Vol. 100, 109596
Effect of cortical bone micro-structure in fragility fracture patients on lamellar stress
- 1 February 2020
- Published by Elsevier in Journal of Biomechanics
- Vol. 100, 109596
Effect of cortical bone micro-structure in fragility fracture patients on lamellar stress