Georges El Fakhri

Georges El Fakhri


Research Focus & Metrics

Academic Career Length

25Years (2002-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
12 December 2023 in Neurology
Verified
Journal Article
Closed access
PDF file
BHBernard Hanseeuw
Bernard Hanseeuw
Heidi I.L. JacobsHeidi I.L. Jacobs
Heidi I.L. Jacobs
ASAaron P. Schultz
Aaron P. Schultz
RBRachel F. Buckley
Rachel F. Buckley
MFMichelle E. Farrell
Michelle E. Farrell
KJKeith A. Johnson
Keith A. Johnson
Background: Hippocampal volume (HV) atrophy is a well-known biomarker of memory impairment. However, compared to amyloid-beta (Aβ) and tau imaging, it is less specific for Alzheimer’s disease (AD) pathology. This lack of specificity could provide indirect information about potential co-pathologies that cannot be observed in vivo. In this prospective cohort study, we aimed to assess the associations among Aβ, tau, HV, and cognition, measured over a ten-year follow-up period with a special focus on the contributions of HV atrophy to cognition after adjusting for Aβ and tau. Methods: We enrolled 283 older adults without dementia or overt cognitive impairment in the Harvard Aging Brain Study. In this report, we only analyzed data from individuals with available longitudinal imaging and cognition data. Serial MRI [follow-up duration: 1.3-7.0y], neocortical Aβ imaging on PiB PET scans [1.9-8.5y], entorhinal and inferior temporal tau on Flortaucipir PET scans [0.8-6.0y], and the Preclinical Alzheimer Cognitive Composite [3.0-9.8y] were prospectively collected. We evaluated the longitudinal associations between Aβ, tau, volume, and cognition data and investigated sequential models to test the contribution of each biomarker to cognitive decline. Results: We analyzed data from 128 clinically normal older adults, including 72 (56%) women and 56 (44%) men; median age at inclusion was 73 years old (range: 63-87). Thirty-four participants (27%) exhibited an initial high-Aβ burden on PET imaging. Faster HV atrophy was correlated with faster cognitive decline (R2=0.28, p<0.0001). When comparing all biomarkers, HV slope was associated with cognitive decline independently of Aβ and tau measures, uniquely accounting for 10% of the variance. Altogether, 45% of the variance in cognitive decline was explained by combining the change measures in the different imaging biomarkers. Discussion: In older adults, longitudinal hippocampal atrophy is associated with cognitive decline, independently of Aβ or tau, suggesting that non-AD pathologies (e.g., TDP-43, vascular) may contribute to hippocampal-mediated cognitive decline. Serial HV measures, in addition to AD-specific biomarkers, may help evaluate the contribution of non-AD pathologies that cannot be measured otherwise in vivo.

23 December 2015 in Journal of Nuclear Medicine
Verified
Journal Article
Open access
PDF file
KGKira S. Grogg
Kira S. Grogg
TTTerrence Toole
Terrence Toole
JOJinsong Ouyang
Jinsong Ouyang
XZXuping Zhu
Xuping Zhu
MNMarc D. Normandin
Marc D. Normandin
GFGeorges El Fakhri
Georges El Fakhri
The aim of this study was to determine the performance of a novel mobile human brain/small-animal PET/CT system. The scanner has a 35.7-cm-diameter bore and a 22-cm axial extent. The detector ring has 7 modules each with 3 × 4 cerium-doped lutetium yttrium orthosilicate crystal blocks, each consisting of 22 × 22 outer-layer and 21 × 21 inner-layer crystals, each layer 1-cm thick. Light is collected by 12 × 12 silicon photomultipliers. The integrated CT can be used for attenuation correction and anatomic localization. The scanner was designed as a low-cost device that nevertheless produces high-quality PET images with the unique capability of battery-powered propulsion, enabling use in many settings. Methods: Spatial resolution, sensitivity, and noise-equivalent counting rate were measured based on the National Electrical Manufacturers Association NU2-2012 procedures. Reconstruction was done with tight energy and timing cuts—400–650 keV and 7 ns—and loose cuts—350–700 keV and 10 ns. Additional image quality measurements were made from phantom, human, and animal studies. Performance was compared with a reference scanner with comparable imaging properties. Results: The full width at half maximum transverse resolution at a 1-cm (10-cm) radius was 3.2 mm (5.2-mm radial, 3.1-mm tangential), and the axial resolution was 3.5 mm (4.0 mm). A sensitivity of 7.5 and 11.7 kcps/MBq at the center for tight and loose cuts, respectively, increased to 8.8 and 13.9 kcps/MBq, respectively, at a 10-cm radial offset. The maximum noise-equivalent counting rate of 19.5 and 22.7 kcps for tight and loose cuts, respectively, was achieved for an activity concentration of 2.9 kBq/mL. Contrast recovery for 4:1 hot cylinder to warm background was 76% for the 25-mm-diameter cylinder but decreased with decreasing cylinder size. The quantitation agreed within 2% of the known activity distribution and concentration. Brain phantom and human scans have shown agreement in SUVs and image quality with the reference scanner. Conclusion: We characterized the performance of the NeuroPET/CT and showed images from the first human studies. The study shows that this scanner achieves good performance when spatial resolution, sensitivity, counting rate, and image quality along with a low cost and unique mobile capabilities are considered.

2 August 2021 in IEEE Journal of Biomedical and Health Informatics
Verified
Journal Article
Closed access
Xiaofeng LiuXiaofeng Liu
Xiaofeng Liu
FXFangxu Xing
Fangxu Xing
CYChao Yang
Chao Yang
CKC.-C. Jay Kuo
C.-C. Jay Kuo
SBSuma Babu
Suma Babu
JWJonghye Woo
Jonghye Woo
Deep learning has great potential for accurate detection and classification of diseases with medical imaging data, but the performance is often limited by the number of training datasets and memory requirements. In addition, many deep learning models are considered a “black-box,” thereby often limiting their adoption in clinical applications. To address this, we present a successive subspace learning model, termed VoxelHop, for accurate classification of Amyotrophic Lateral Sclerosis (ALS) using T2-weighted structural MRI data. Compared with popular convolutional neural network (CNN) architectures, VoxelHop has modular and transparent structures with fewer parameters without any backpropagation, so it is well-suited to small dataset size and 3D imaging data. Our VoxelHop has four key components, including (1) sequential expansion of near-to-far neighborhood for multi-channel 3D data; (2) subspace approximation for unsupervised dimension reduction; (3) label-assisted regression for supervised dimension reduction; and (4) concatenation of features and classification between controls and patients. Our experimental results demonstrate that our framework using a total of 20 controls and 26 patients achieves an accuracy of 93.48 $\%$ and an AUC score of 0.9394 in differentiating patients from controls, even with a relatively small number of datasets, showing its robustness and effectiveness. Our thorough evaluations also show its validity and superiority to the state-of-the-art 3D CNN classification approaches. Our framework can easily be generalized to other classification tasks using different imaging modalities.

1 July 2015 in British Journal of Radiology
Verified
Journal Article
Open access
HPHarald Paganetti
Harald Paganetti
GFGeorges El Fakhri
Georges El Fakhri
Protons are being used in radiation therapy because of typically better dose conformity and reduced total energy deposited in the patient as compared with photon techniques. Both aspects are related to the finite range of a proton beam. The finite range also allows advanced dose shaping. These benefits can only be fully utilized if the end of range can be predicted accurately in the patient. The prediction of the range in tissue is associated with considerable uncertainties owing to imaging, patient set-up, beam delivery, interfractional changes in patient anatomy and dose calculation. Consequently, a significant range (of the order of several millimetres) is added to the prescribed range in order to ensure tumour coverage. Thus, reducing range uncertainties would allow a reduction of the treatment volume and reduce dose to potential organs at risk.

1 February 2020 in International Journal of Nanomedicine
Verified
Journal Article
Open access
PDF file
HGHaiyan Gao
Haiyan Gao
YBYan Bai
Yan Bai
LCLijuan Chen
Lijuan Chen
GFGeorges El Fakhri
Georges El Fakhri
MWMeiyun Wang
Meiyun Wang
Background and Purpose: The development of multiple drug resistance (MDR) to chemotherapy and single modal therapy remains unsatisfied for the eradication of tumor, which are major obstacles in cancer therapy. This novel system with excellent characteristics for inhibition of P-glycoprotein (P-gp), and for near-infrared fluorescence (NIRF) imaging-guided chemo-photothermal therapy (PTT), has been identified as a promising way to MDR and achieve synergistic cancer therapy. Methods: In this study, we successfully synthesized a multifunctional theranostic system, which was developed through FDA-approved self-assembling drugs, which contain anticancer drug doxorubicin (Dox), imaging and high photothermal conversion drug indocyanine green (ICG) and P-gp regulator TPGS (the system named T/Dox-ICG). We studied the characterization of T/Dox-ICG NPs, including the TEM, SEM, DLS, UV-vis-NIR, zeta potential, CLSM, in vitro FL imaging, in vitro photothermal effect, in vitro Dox and ICG release. We used CLSM to verify the location of intracellular distribution of Dox in SCG 7901/VCR cells, Western blot was performed to demonstrate the TPGS-mediated inhibition of P-gp. And, the cytotoxicity of materials against SCG 7901/VCR cells was studied by the MTT assay. Results: The TEM showed the T/Dox-ICG NPs had good monodispersity with diameters of 19.03 nm, Dox and ICG could be released constantly from T/Dox-ICG NPs in vitro. In vitro cell experiments demonstrated higher Dox accumulation and retention in the nucleus. Western blot showed TPGS could obviously inhibit the expression of P-gp. In vitro cytotoxicity assay showed more significant cytotoxicity on MDR cells (SCG 7901/VCR) with only 8.75% of cells surviving. Conclusion: MDR cancer therapy indicates that it may be important to develop a safer system that can simultaneously inhibit the drug transporters and monitor the delivery of chemotherapeutic agents, and combination therapy have raised widespread concern on tumor treatment.

9 January 2024 in Neurology
Verified
Journal Article
Closed access
ABAndrée-Ann Baril
Andrée-Ann Baril
DKDaniel J. Kojis
Daniel J. Kojis
JHJayandra J. Himali
Jayandra J. Himali
CDCharles DeCarli
Charles DeCarli
ESErlan Sanchez
Erlan Sanchez
SSSudha Seshadri
Sudha Seshadri
Both short and long sleep duration were previously associated with incident dementia, but underlying mechanisms remain unclear. We evaluated how self-reported sleep duration and its change over time associate with (A)myloid, (T)au, (N)eurodegeneration, and (V)ascular neuroimaging markers of Alzheimer disease. Two Framingham Heart Study overlapping samples were studied: participants who underwent 11C-Pittsburg Compound B amyloid and 18F-flortaucipir tau PET imaging and participants who underwent an MRI. MRI metrics estimated neurodegeneration (total brain volume) and cerebrovascular injuries (white matter hyperintensities [WMHs] volume, covert brain infarcts, free-water [FW] fraction). Self-reported sleep duration was assessed and split into categories both at the time of neuroimaging testing and approximately 13 years before: short ≤6 hours. average 7-8 hours, and long ≥9 hours. Logistic and linear regression models were used to examine sleep duration and neuroimaging metrics. The tested cohort was composed of 271 participants (age 53.6 ± 8.0 years; 51% male) in the PET imaging sample and 2,165 participants (age 61.3 ± 11.1 years; 45% male) in the MRI sample. No fully adjusted association was observed between cross-sectional sleep duration and neuroimaging metrics. In fully adjusted models compared with consistently sleeping 7-8 hours, groups transitioning to a longer sleep duration category over time had higher FW fraction (short to average β [SE] 0.0062 [0.0024], p = 0.009; short to long β [SE] 0.0164 [0.0076], p = 0.031; average to long β [SE] 0.0083 [0.0022], p = 0.002), and those specifically going from average to long sleep duration also had higher WMH burden (β [SE] 0.29 [0.11], p = 0.007). The opposite associations (lower WMH and FW) were observed in participants consistently sleeping ≥9 hours as compared with people consistently sleeping 7-8 hours in fully adjusted models (β [SE] -0.43 [0.20], p = 0.028; β [SE] -0.019 [0.004], p = 0.020). Each hour of increasing sleep (continuous, β [SE] 0.12 [0.04], p = 0.003; β [SE] 0.002 [0.001], p = 0.021) and extensive increase in sleep duration (≥2 hours vs 0 ± 1 hour change; β [SE] 0.24 [0.10], p = 0.019; β [SE] 0.0081 [0.0025], p = 0.001) over time was associated with higher WMH burden and FW fraction in fully adjusted models. Sleep duration change was not associated with PET amyloid or tau outcomes. Longer self-reported sleep duration over time was associated with neuroimaging biomarkers of cerebrovascular pathology as evidenced by higher WMH burden and FW fraction. A longer sleep duration extending over time may be an early change in the neurodegenerative trajectory.