Anousheh Sayah

Dr. Anousheh Sayah

orcid

Research Focus & Metrics

Academic Career Length

12Years (2015-Present)

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

Representative Publications

View All Publications
5 December 2022 in Nature Communications
Verified
Journal Article
Open access
PDF file
SPSarthak Pati
Sarthak Pati
Ujjwal BaidUjjwal Baid
Ujjwal Baid
BEBrandon Edwards
Brandon Edwards
MSMicah Sheller
Micah Sheller
SWShih-Han Wang
Shih-Han Wang
Spyridon BakasSpyridon Bakas
Spyridon Bakas
Although machine learning (ML) has shown promise across disciplines, out-of-sample generalizability is concerning. This is currently addressed by sharing multi-site data, but such centralization is challenging/infeasible to scale due to various limitations. Federated ML (FL) provides an alternative paradigm for accurate and generalizable ML, by only sharing numerical model updates. Here we present the largest FL study to-date, involving data from 71 sites across 6 continents, to generate an automatic tumor boundary detector for the rare disease of glioblastoma, reporting the largest such dataset in the literature (n = 6, 314). We demonstrate a 33% delineation improvement for the surgically targetable tumor, and 23% for the complete tumor extent, over a publicly trained model. We anticipate our study to: 1) enable more healthcare studies informed by large diverse data, ensuring meaningful results for rare diseases and underrepresented populations, 2) facilitate further analyses for glioblastoma by releasing our consensus model, and 3) demonstrate the FL effectiveness at such scale and task-complexity as a paradigm shift for multi-site collaborations, alleviating the need for data-sharing.

8 July 2025 in Nature Communications
Verified
Journal Article
Open access
PDF file
MZMaximilian Zenk
Maximilian Zenk
Ujjwal BaidUjjwal Baid
Ujjwal Baid
SPSarthak Pati
Sarthak Pati
ALAkis Linardos
Akis Linardos
BEBrandon Edwards
Brandon Edwards
Spyridon BakasSpyridon Bakas
Spyridon Bakas
Computational competitions are the standard for benchmarking medical image analysis algorithms, but they typically use small curated test datasets acquired at a few centers, leaving a gap to the reality of diverse multicentric patient data. To this end, the Federated Tumor Segmentation (FeTS) Challenge represents the paradigm for real-world algorithmic performance evaluation. The FeTS challenge is a competition to benchmark (i) federated learning aggregation algorithms and (ii) state-of-the-art segmentation algorithms, across multiple international sites. Weight aggregation and client selection techniques were compared using a multicentric brain tumor dataset in realistic federated learning simulations, yielding benefits for adaptive weight aggregation, and efficiency gains through client sampling. Quantitative performance evaluation of state-of-the-art segmentation algorithms on data distributed internationally across 32 institutions yielded good generalization on average, albeit the worst-case performance revealed data-specific modes of failure. Similar multi-site setups can help validate the real-world utility of healthcare AI algorithms in the future.

1 September 2016 in American Journal of Roentgenology
Verified
Journal Article
Closed access
ASAnousheh Sayah
Anousheh Sayah
AJAnn K. Jay
Ann K. Jay
JTJacob S. Toaff
Jacob S. Toaff
EMErini V. Makariou
Erini V. Makariou
FBFrank Berkowitz
Frank Berkowitz
OBJECTIVE. Reducing lumbar spine MRI scanning time while retaining diagnostic accuracy can benefit patients and reduce health care costs. This study compares the effectiveness of a rapid lumbar MRI protocol using 3D T2-weighted sampling perfection with application-optimized contrast with different flip-angle evolutions (SPACE) sequences with a standard MRI protocol for evaluation of lumbar spondylosis. MATERIALS AND METHODS. Two hundred fifty consecutive unenhanced lumbar MRI examinations performed at 1.5 T were retrospectively reviewed. Full, rapid, and complete versions of each examination were interpreted for spondylotic changes at each lumbar level, including herniations and neural compromise. The full examination consisted of sagittal T1-weighted, T2-weighted turbo spin-echo (TSE), and STIR sequences; and axial T1- and T2-weighted TSE sequences (time, 18 minutes 40 seconds). The rapid examination consisted of sagittal T1- and T2-weighted SPACE sequences, with axial SPACE reformations (time, 8 minutes...

1 November 2015 in The Spine Journal
Verified
Journal Article
Closed access
LKLinda C. Kelahan
Linda C. Kelahan
FSFaheem A. Sandhu
Faheem A. Sandhu
ASAnousheh Sayah
Anousheh Sayah

29 October 2015 in American Journal of Neuroradiology
Verified
Journal Article
Open access
PDF file
MBM.P. Bazylewicz
M.P. Bazylewicz
FBF. Berkowitz
F. Berkowitz
ASAnousheh Sayah
Anousheh Sayah
BACKGROUND AND PURPOSE: The current standard technique for cervical puncture involves prone positioning with neck extension. The purpose of this study was to compare measurements of the posterior cervical thecal sac during neck flexion and extension in supine and prone positions by using high-resolution MR imaging to help determine the optimal positioning for cervical puncture. MATERIALS AND METHODS: High-resolution T2-weighted MR imaging was performed of the cervical spine in 10 adult volunteers 18 years of age and older. Exclusion criteria included the following: a history of cervical spine injury/surgery, neck pain, and degenerative spondylosis. Images of sagittal 3D sampling perfection with application-optimized contrasts by using different flip angle evolutions were obtained in the following neck positions: supine extension, supine flexion, prone extension, and prone flexion. The degree of neck flexion and extension and the distance from the posterior margin of the spinal cord to the posterior aspect of the C1–C2 thecal sac were measured in each position. RESULTS: The mean anteroposterior size of the posterior C1–C2 thecal sac was as follows: 4.76 mm for supine extension, 3.63 mm for supine flexion, 5.00 mm for prone extension, and 4.00 mm for prone flexion. Neck extension yielded a larger CSF space than flexion, independent of supine/prone positioning. There was no correlation with neck angle and thecal sac size. CONCLUSIONS: The posterior C1–C2 thecal sac is larger with neck extension than flexion, independent of prone or supine positioning. Given that this space is the target for cervical puncture, findings suggest that extension is the ideal position for performing the procedure, and the decision for prone-versus-supine positioning can be made on the basis of operator comfort and patient preference/ability.

8 March 2021 in The Neuroradiology Journal
Verified
Journal Article
Closed access
PDF file
ASAnousheh Sayah
Anousheh Sayah
AFApril D Farley
April D Farley
EMEric C Munoz
Eric C Munoz
FSFaheem A Sandhu
Faheem A Sandhu
FBFrank Berkowitz
Frank Berkowitz
An abnormally decreased clivoaxial angle (CXA) is used during the clinical evaluation for corrective skull base surgery. Published normal ranges of CXA using x-ray, computed tomography, or magnetic resonance imaging (MRI) vary dramatically, especially with neck flexion or extension. The aim of this study was to use high-resolution MRI to determine the normal range of CXA in various neck positions using a reproducible measurement technique. The CXA was measured in 10 healthy volunteers on sagittal T2 SPACE c-spine MRI in supine and prone positions and with the neck both neck and extended. CXA is strictly defined as the angle between a line along the inferior third of the dorsal clival cortex and a line from the superior/posterior cortex of the dens to the posterior/inferior corner of the C2 body. Statistical analysis was performed in all positions and included mean CXA, range, standard deviation (SD), inter-reader agreement, and group comparisons. The mean CXA overall was 156.92° (SD=4.23°; range 134–179°). The mean value for extension CXA was 169.20° (SD=5.81°), and the mean value for flexion CXA was 144.73° (SD=5.71°), the difference being statistically significant (p<0.0001) regardless of supine or prone position. Concordant correlations of reader measurements showed substantial agreement in the supine position at 0.96, with lower agreement in the prone position at 0.87. We report normal ranges for CXA in various neck positions based on 3D T2-weighted MRI, using a reproducible measurement method. There was a significant difference in the CXA values between neck extended and neck flexed positions but not between supine and prone positions.