Nergis C. Khan
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
Nergis C. Khan has a total of 11 co-authors. Here are the top co-authors by number of shared publications:
- Anoopum S. Gupta
Department of Neurology, Massachusetts General Hospital, Harvard Medical School, Boston, USA
1 shared publication - Chandrashan Perera
Department of Ophthalmology, Stanford University School of Medicine, Stanford, CA 94305, USA
1 shared publication - David Myung
Stanford University School of Medicine
1 shared publication - Diana V. Do
Byers Eye Institute, Stanford University School of Medicine, Palo Alto, USA
1 shared publication - Eliot R. Dow
Byers Eye Institute at Stanford, Department of Ophthalmology, Stanford University School of Medicine, Palo Alto, CA 94305, USA
1 shared publication - Karen M. Chen
Medical Student, Columbia University Vagelos College of Physicians and Surgeons
1 shared publication
Latest Publications
- Open Access
- 14 July 2022
- Published by MDPI AG in Diagnostics
- Vol. 12 (7) , 1714
Predicting Systemic Health Features from Retinal Fundus Images Using Transfer-Learning-Based Artificial Intelligence Models
While color fundus photos are used in routine clinical practice to diagnose ophthalmic conditions, evidence suggests that ocular imaging contains valuable information regarding the systemic health features of patients. These features can be identified through computer vision techniques including deep learning (DL) artificial intelligence (AI) models. We aim to construct a DL model that can predict systemic features from fundus images and to determine the optimal method of model construction for this task. Data were collected from a cohort of patients undergoing diabetic retinopathy screening between March 2020 and March 2021. Two models were created for each of 12 systemic health features based on the DenseNet201 architecture: one utilizing transfer learning with images from ImageNet and another from 35,126 fundus images. Here, 1277 fundus images were used to train the AI models. Area under the receiver operating characteristics curve (AUROC) scores were used to compare the model performance. Models utilizing the ImageNet transfer learning data were superior to those using retinal images for transfer learning (mean AUROC 0.78 vs. 0.65, p-value < 0.001). Models using ImageNet pretraining were able to predict systemic features including ethnicity (AUROC 0.93), age > 70 (AUROC 0.90), gender (AUROC 0.85), ACE inhibitor (AUROC 0.82), and ARB medication use (AUROC 0.78). We conclude that fundus images contain valuable information about the systemic characteristics of a patient. To optimize DL model performance, we recommend that even domain specific models consider using transfer learning from more generalized image sets to improve accuracy.
- 14 July 2022
- Published by MDPI AG in Diagnostics
- Vol. 12 (7) , 1714
Open AccessPredicting Systemic Health Features from Retinal Fundus Images Using Transfer-Learning-Based Artificial Intelligence Models
While color fundus photos are used in routine clinical practice to diagnose ophthalmic conditions, evidence suggests that ocular imaging contains valuable information regarding the systemic health features of patients. These features can be identified through computer vision techniques including deep learning (DL) artificial intelligence (AI) models. We aim to construct a DL model that can predict systemic features from fundus images and to determine the optimal method of model construction for this task. Data were collected from a cohort of patients undergoing diabetic retinopathy screening between March 2020 and March 2021. Two models were created for each of 12 systemic health features based on the DenseNet201 architecture: one utilizing transfer learning with images from ImageNet and another from 35,126 fundus images. Here, 1277 fundus images were used to train the AI models. Area under the receiver operating characteristics curve (AUROC) scores were used to compare the model performance. Models utilizing the ImageNet transfer learning data were superior to those using retinal images for transfer learning (mean AUROC 0.78 vs. 0.65, p-value < 0.001). Models using ImageNet pretraining were able to predict systemic features including ethnicity (AUROC 0.93), age > 70 (AUROC 0.90), gender (AUROC 0.85), ACE inhibitor (AUROC 0.82), and ARB medication use (AUROC 0.78). We conclude that fundus images contain valuable information about the systemic characteristics of a patient. To optimize DL model performance, we recommend that even domain specific models consider using transfer learning from more generalized image sets to improve accuracy.
- 24 July 2021
- Published by Springer Nature in The Cerebellum
- Vol. 21 (3) , 368-379
Free-Living Motor Activity Monitoring in Ataxia-Telangiectasia
With disease-modifying approaches under evaluation in ataxia-telangiectasia and other ataxias, there is a need for objective and reliable biomarkers of free-living motor function. In this study, we test the hypothesis that metrics derived from a single wrist sensor worn at home provide accurate, reliable, and interpretable information about neurological disease severity in children with A-T. A total of 15 children with A-T and 15 age- and sex-matched controls wore a sensor with a triaxial accelerometer on their dominant wrist for 1 week at home. Activity intensity measures, derived from the sensor data, were compared with in-person neurological evaluation on the Brief Ataxia Rating Scale (BARS) and performance on a validated computer mouse task. Children with A-T were inactive the same proportion of each day as controls but produced more low intensity movements (p < 0.01; Cohen’s d = 1.48) and fewer high intensity movements (p < 0.001; Cohen’s d = 1.71). The range of activity intensities was markedly reduced in A-T compared to controls (p < 0.0001; Cohen’s d = 2.72). The activity metrics correlated strongly with arm, gait, and total clinical severity (r: 0.71–0.87; p < 0.0001), correlated with specific computer task motor features (r: 0.67–0.92; p < 0.01), demonstrated high reliability (r: 0.86–0.93; p < 0.00001), and were not significantly influenced by age in the healthy control group. Motor activity metrics from a single, inexpensive wrist sensor during free-living behavior provide accurate and reliable information about diagnosis, neurological disease severity, and motor performance. These low-burden measurements are applicable independent of ambulatory status and are potential digital behavioral biomarkers in A-T.
- 24 July 2021
- Published by Springer Nature in The Cerebellum
- Vol. 21 (3) , 368-379
Free-Living Motor Activity Monitoring in Ataxia-Telangiectasia
With disease-modifying approaches under evaluation in ataxia-telangiectasia and other ataxias, there is a need for objective and reliable biomarkers of free-living motor function. In this study, we test the hypothesis that metrics derived from a single wrist sensor worn at home provide accurate, reliable, and interpretable information about neurological disease severity in children with A-T. A total of 15 children with A-T and 15 age- and sex-matched controls wore a sensor with a triaxial accelerometer on their dominant wrist for 1 week at home. Activity intensity measures, derived from the sensor data, were compared with in-person neurological evaluation on the Brief Ataxia Rating Scale (BARS) and performance on a validated computer mouse task. Children with A-T were inactive the same proportion of each day as controls but produced more low intensity movements (p < 0.01; Cohen’s d = 1.48) and fewer high intensity movements (p < 0.001; Cohen’s d = 1.71). The range of activity intensities was markedly reduced in A-T compared to controls (p < 0.0001; Cohen’s d = 2.72). The activity metrics correlated strongly with arm, gait, and total clinical severity (r: 0.71–0.87; p < 0.0001), correlated with specific computer task motor features (r: 0.67–0.92; p < 0.01), demonstrated high reliability (r: 0.86–0.93; p < 0.00001), and were not significantly influenced by age in the healthy control group. Motor activity metrics from a single, inexpensive wrist sensor during free-living behavior provide accurate and reliable information about diagnosis, neurological disease severity, and motor performance. These low-burden measurements are applicable independent of ambulatory status and are potential digital behavioral biomarkers in A-T.