Mark J. Kupersmith
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Mark J. Kupersmith has a total of 143 co-authors. Here are the top co-authors by number of shared publications:
Tobias Elze
Schepens Eye Research Institute, Harvard Medical School, Boston, USA
9 shared publicationsMichael Wall
Department of Ophthalmology and Visual Sciences, University of Iowa, Iowa City, Iowa, United States
7 shared publicationsDavid Szanto
Department of Ophthalmology), New York Eye and Ear Infirmary at Mount Sinai, New York, USA
6 shared publicationsBrian Woods
Department of Ophthalmology), Irish Clinical Academic Training Programme , University Hospital Galway, Galway, Ireland
5 shared publicationsJoseph Branco
New York Medical College, Valhalla, USA
5 shared publicationsJui-Kai Wang
Department of Ophthalmology, University of Texas Southwestern Medical Center, Dallas, Texas, United States
5 shared publications
Latest Publications
- Open Access
- 23 September 2026
- Published by Wolters Kluwer Health in Journal of Neuro-Ophthalmology
Optic Neuritis Model to Identify Potential Neuroprotective Therapy
Was not provided for this article.
- 23 September 2026
- Published by Wolters Kluwer Health in Journal of Neuro-Ophthalmology
Open AccessOptic Neuritis Model to Identify Potential Neuroprotective Therapy
Was not provided for this article.
- 18 September 2026
- Published by Elsevier in Ophthalmology
- Vol. 133 (11) , 1306-1307
Nonarteritic Anterior Ischemic Optic Neuropathy and Glucagon-Like Peptide-1 Agonist Therapies: An Important Rare Issue, Lacking Adequate Data
- 18 September 2026
- Published by Elsevier in Ophthalmology
- Vol. 133 (11) , 1306-1307
Nonarteritic Anterior Ischemic Optic Neuropathy and Glucagon-Like Peptide-1 Agonist Therapies: An Important Rare Issue, Lacking Adequate Data
- Open Access
- 5 January 2026
- Published by Association for Research in Vision and Ophthalmology (ARVO) in Investigative Ophthalmology & Visual Science
- Vol. 67 (1) , 12
Multimodal Deep Learning Differentiates Papilledema and Non-Arteritic Anterior Ischemic Optic Neuropathy From Healthy Eyes
Optic nerve head (ONH) swelling, a critical feature in idiopathic intracranial hypertension (IIH) and non-arteritic anterior ischemic optic neuropathy (NAION), can present diagnostic challenges. We explored a multimodal deep learning (DL) approach integrating optical coherence tomography (OCT) scans and fundus photographs to enhance diagnostic accuracy for differentiating IIH, NAION, and healthy eyes. We developed two separate models using 7019 OCT scans (3D-ResNet-18) and 17,657 fundus photos (ResNet-50) to classify eyes with papilledema (2315 OCT, 6349 fundus), NAION (841 OCT, 1814 fundus), and healthy eyes (3863 OCT, 9494 fundus). We arranged the dataset so that the test set consisted entirely of same-day OCT scans and fundus photos, with each modality (OCT and fundus) contributing at least 15% of the data for each class. We combined output probabilities from both models using two methods: an F1-weighted sum by class (F1WS), as well as an XGBoost model. Performance of each was evaluated with AUC-ROC, accuracy, precision, recall, and F1 scores. The OCT model alone achieved a test accuracy of 93.5%, with the fundus photo model reaching 93.9%. The multimodal F1WS and XGBoost models achieved an accuracy of 97.5% and 98.3%, respectively. Combining OCT and fundus photographs improves the classification of IIH, NAION, and healthy eyes, showing the value of using complementary imaging modalities. This approach supports the use of DL to aid diagnosis and clinical management of optic nerve head swelling. It may also be extended to leverage DL from additional data sources, such as macular scans or visual field tests.
- 5 January 2026
- Published by Association for Research in Vision and Ophthalmology (ARVO) in Investigative Ophthalmology & Visual Science
- Vol. 67 (1) , 12
Open AccessMultimodal Deep Learning Differentiates Papilledema and Non-Arteritic Anterior Ischemic Optic Neuropathy From Healthy Eyes
Optic nerve head (ONH) swelling, a critical feature in idiopathic intracranial hypertension (IIH) and non-arteritic anterior ischemic optic neuropathy (NAION), can present diagnostic challenges. We explored a multimodal deep learning (DL) approach integrating optical coherence tomography (OCT) scans and fundus photographs to enhance diagnostic accuracy for differentiating IIH, NAION, and healthy eyes. We developed two separate models using 7019 OCT scans (3D-ResNet-18) and 17,657 fundus photos (ResNet-50) to classify eyes with papilledema (2315 OCT, 6349 fundus), NAION (841 OCT, 1814 fundus), and healthy eyes (3863 OCT, 9494 fundus). We arranged the dataset so that the test set consisted entirely of same-day OCT scans and fundus photos, with each modality (OCT and fundus) contributing at least 15% of the data for each class. We combined output probabilities from both models using two methods: an F1-weighted sum by class (F1WS), as well as an XGBoost model. Performance of each was evaluated with AUC-ROC, accuracy, precision, recall, and F1 scores. The OCT model alone achieved a test accuracy of 93.5%, with the fundus photo model reaching 93.9%. The multimodal F1WS and XGBoost models achieved an accuracy of 97.5% and 98.3%, respectively. Combining OCT and fundus photographs improves the classification of IIH, NAION, and healthy eyes, showing the value of using complementary imaging modalities. This approach supports the use of DL to aid diagnosis and clinical management of optic nerve head swelling. It may also be extended to leverage DL from additional data sources, such as macular scans or visual field tests.
- 25 December 2025
- Published by BMJ in British Journal of Ophthalmology
- Vol. 110 (7) , 736-742
Archetypal analysis quantifies changes in patterns of visual field loss in eyes with acute non-arteritic anterior ischaemic optic neuropathy
Aims: We characterised visual field (VF) spatial loss patterns in acute non-arteritic anterior ischaemic optic neuropathy (NAION) using archetypal analysis (AA). Methods: We performed standard automated perimetry on 727 participants with acute NAION at screening, day 1 of enrolment, months 2, 6 and 12. We applied AA, an unsupervised machine learning technique, to identify and quantify distinct VF loss patterns (archetypes, ATs). We used Mann-Whitney U tests and Wilcoxon signed-rank tests to assess demographic differences and longitudinal changes in discrete variables, and linear regression to analyse continuous variables. Results: AA identified 10 distinct ATs, with three especially prevalent patterns: global VF loss (AT1), inferior altitudinal loss (AT2) and inferior altitudinal with macular sparing (AT3). Between day 1 and month 2, AT1 RW significantly increased from 15.7% to 28.1% (false discovery rate, FDR-adjusted p<0.001). Asian participants consistently exhibited greater RW for AT1 compared with White participants (day 1: 40.7% vs 17.7%; FDR-adjusted p<0.001). Sex differences emerged modestly at months 2 and 6, with females having a higher RW for superior altitudinal loss (FDR-adjusted p=0.049). Older participants showed slightly greater frequency of central horizontal and mild central depression patterns (FDR-adjusted p<0.001). Conclusions: AA effectively quantifies distinct, clinically significant VF spatial loss patterns in NAION, revealing significant temporal changes and demographic differences. Global VF loss represents the predominant AT, increasing notably within 2 months of disease onset. Prominent racial disparities, particularly higher severity in Asian individuals, underscore potential differences in NAION aetiology or susceptibility. These findings provide a foundation for improved disease characterisation and prognosis.
- 25 December 2025
- Published by BMJ in British Journal of Ophthalmology
- Vol. 110 (7) , 736-742
Archetypal analysis quantifies changes in patterns of visual field loss in eyes with acute non-arteritic anterior ischaemic optic neuropathy
Aims: We characterised visual field (VF) spatial loss patterns in acute non-arteritic anterior ischaemic optic neuropathy (NAION) using archetypal analysis (AA). Methods: We performed standard automated perimetry on 727 participants with acute NAION at screening, day 1 of enrolment, months 2, 6 and 12. We applied AA, an unsupervised machine learning technique, to identify and quantify distinct VF loss patterns (archetypes, ATs). We used Mann-Whitney U tests and Wilcoxon signed-rank tests to assess demographic differences and longitudinal changes in discrete variables, and linear regression to analyse continuous variables. Results: AA identified 10 distinct ATs, with three especially prevalent patterns: global VF loss (AT1), inferior altitudinal loss (AT2) and inferior altitudinal with macular sparing (AT3). Between day 1 and month 2, AT1 RW significantly increased from 15.7% to 28.1% (false discovery rate, FDR-adjusted p<0.001). Asian participants consistently exhibited greater RW for AT1 compared with White participants (day 1: 40.7% vs 17.7%; FDR-adjusted p<0.001). Sex differences emerged modestly at months 2 and 6, with females having a higher RW for superior altitudinal loss (FDR-adjusted p=0.049). Older participants showed slightly greater frequency of central horizontal and mild central depression patterns (FDR-adjusted p<0.001). Conclusions: AA effectively quantifies distinct, clinically significant VF spatial loss patterns in NAION, revealing significant temporal changes and demographic differences. Global VF loss represents the predominant AT, increasing notably within 2 months of disease onset. Prominent racial disparities, particularly higher severity in Asian individuals, underscore potential differences in NAION aetiology or susceptibility. These findings provide a foundation for improved disease characterisation and prognosis.
- Open Access
- 3 November 2025
- Published by Frontiers Media SA in Frontiers in Neurology
- Vol. 16, 1678151
Ocular myasthenia gravis management
Ocular myasthenia gravis (OMG) has no agreed-upon diagnostic and management criteria, leading to misdiagnoses and often misguided treatments. The purpose of this minireview is to provide guidance to clinicians who encounter possible OMG patients. We cite relevant literature and the recommendations for diagnosis and therapy based on the authors’ extensive experience in OMG. This report provides sound recommendations based on the authors’ successes and failures, coupled with relevant literature in the myasthenia gravis field.
- 3 November 2025
- Published by Frontiers Media SA in Frontiers in Neurology
- Vol. 16, 1678151
Open AccessOcular myasthenia gravis management
Ocular myasthenia gravis (OMG) has no agreed-upon diagnostic and management criteria, leading to misdiagnoses and often misguided treatments. The purpose of this minireview is to provide guidance to clinicians who encounter possible OMG patients. We cite relevant literature and the recommendations for diagnosis and therapy based on the authors’ extensive experience in OMG. This report provides sound recommendations based on the authors’ successes and failures, coupled with relevant literature in the myasthenia gravis field.
- Open Access
Predicting Acute Cerebrovascular Events in Stroke Alerts Using Large-Language Models and Structured Data
Background Acute stroke alerts are often activated for non-cerebrovascular conditions, leading to false positives that strain clinical resources and promote diagnostic uncertainty. We sought to develop machine learning (ML) models integrating large-language models (LLMs), structured electronic health record data, and clinical time series data to predict the presence of acute cerebrovascular disease (ACD) at stroke alert activation. Methods We derived a series of ML models using retrospective data from stroke alerts activated at Mount Sinai Health System between 2011 and 2021. We extracted structured data (demographics, medical comorbidities, medications, and engineered time-series features from vital signs and lab results) as well as unstructured clinical notes available prior to the time of stroke alert. We processed clinical notes using three embedding approaches: word embeddngs, biomedical embeddings (BioWordVec), and LLMs. Using a radiographic gold standard for acute intracranial vascular event, we used an auto-ML approach to train one model based on unstructured data and five models based on different combinations of structured data. We evaluated models individually using the area under the receiver operating characteristic curve (AUROC), mean positive predictive value (PPV), sensitivity, and F1-score. We then combined the 6 model logits into a multimodal ensemble by weighting their logits based on F1-score, determining ensemble performance using the same metrics. Results We identified 16,512 stroke alerts corresponding to 14,233 unique patients over the study period, of which 9,013 (54.6%) were due to ACD. The multi-modal model (AUROC 0.72, PPV 0.68, sensitivity 0.76, F1 0.72) outperformed all individual models by AUROC. One structured model based on demographics, comorbidities, and medications demonstrated the highest sensitivity (0.95). Conclusions We developed a multi-modal ML model to predict ACD at stroke alert activation. This approach has promise to optimize stroke triage and reduce false-positive activations.
Open AccessPredicting Acute Cerebrovascular Events in Stroke Alerts Using Large-Language Models and Structured Data
Background Acute stroke alerts are often activated for non-cerebrovascular conditions, leading to false positives that strain clinical resources and promote diagnostic uncertainty. We sought to develop machine learning (ML) models integrating large-language models (LLMs), structured electronic health record data, and clinical time series data to predict the presence of acute cerebrovascular disease (ACD) at stroke alert activation. Methods We derived a series of ML models using retrospective data from stroke alerts activated at Mount Sinai Health System between 2011 and 2021. We extracted structured data (demographics, medical comorbidities, medications, and engineered time-series features from vital signs and lab results) as well as unstructured clinical notes available prior to the time of stroke alert. We processed clinical notes using three embedding approaches: word embeddngs, biomedical embeddings (BioWordVec), and LLMs. Using a radiographic gold standard for acute intracranial vascular event, we used an auto-ML approach to train one model based on unstructured data and five models based on different combinations of structured data. We evaluated models individually using the area under the receiver operating characteristic curve (AUROC), mean positive predictive value (PPV), sensitivity, and F1-score. We then combined the 6 model logits into a multimodal ensemble by weighting their logits based on F1-score, determining ensemble performance using the same metrics. Results We identified 16,512 stroke alerts corresponding to 14,233 unique patients over the study period, of which 9,013 (54.6%) were due to ACD. The multi-modal model (AUROC 0.72, PPV 0.68, sensitivity 0.76, F1 0.72) outperformed all individual models by AUROC. One structured model based on demographics, comorbidities, and medications demonstrated the highest sensitivity (0.95). Conclusions We developed a multi-modal ML model to predict ACD at stroke alert activation. This approach has promise to optimize stroke triage and reduce false-positive activations.
- 28 May 2025
- Published by Elsevier in American Journal of Ophthalmology
- Vol. 277, 249-259
Deep Learning Differentiates Papilledema, NAION, and Healthy Eyes With Unsegmented 3D OCT Volumes
- 28 May 2025
- Published by Elsevier in American Journal of Ophthalmology
- Vol. 277, 249-259
Deep Learning Differentiates Papilledema, NAION, and Healthy Eyes With Unsegmented 3D OCT Volumes
- Open Access
- 24 April 2025
- Published by Public Library of Science (PLoS) in PLOS ONE
- Vol. 20 (4) , e0320759
The reliability of medical illness reporting in a randomized clinical trial
Background/Objective: Reported medical disorders from population surveys, medical records, and clinical trials, may not be accurate and methods are needed to improve confirmation. We report the accuracy of reported prevalence of medical disorders in a clinical trial and comparison with potential verification methods. Methods: We report the prevalence of 11 medical disorders, utilizing prospectively collected data from 729 participants in an eight-country multicenter clinical treatment trial on non-arteritic anterior ischemic optic neuropathy (NAION). We chose disorders where the medical history was potentially verifiable. We determined the prevalence using four methods: Method (M)1: Participant and medical health record reporting; M2: Physical examination, clinical tests; M3: Medication indications; M4: Combining M2 and M3. We estimated concordance between M1 and the other methods using Cohen’s kappa (K) statistic. Results: Prevalence of the medical disorders based on M1 were lower than for either M2 or M3, depending on the disorder, and consistentlly lower for M4. For M1 and M4, moderate concordance (K ≥ 0.50) was observed only for psychiatric disorders (K = 0.52) and prior NAION (K=0.67). The prevalence and concordance for M1 and M4 for anemia, hypertension, diabetes and psychiatric disease were the only disorders that differed between females and males. For all methods, the prevalence varied widely across countries. Concordance for M1 and M4 varied and moderate concordance occurred for psychiatric disorders and prior NAION. Conclusion: Even with prospective, rigorously collected data, medical histories do not reliably identify all medical disorders. Adding the results of physical examination, laboratory tests, and medications increases the accuracy of reporting. This strategy could be adapted for clinical trials and electronic medical record disease-prevalence data mining.
- 24 April 2025
- Published by Public Library of Science (PLoS) in PLOS ONE
- Vol. 20 (4) , e0320759
Open AccessThe reliability of medical illness reporting in a randomized clinical trial
Background/Objective: Reported medical disorders from population surveys, medical records, and clinical trials, may not be accurate and methods are needed to improve confirmation. We report the accuracy of reported prevalence of medical disorders in a clinical trial and comparison with potential verification methods. Methods: We report the prevalence of 11 medical disorders, utilizing prospectively collected data from 729 participants in an eight-country multicenter clinical treatment trial on non-arteritic anterior ischemic optic neuropathy (NAION). We chose disorders where the medical history was potentially verifiable. We determined the prevalence using four methods: Method (M)1: Participant and medical health record reporting; M2: Physical examination, clinical tests; M3: Medication indications; M4: Combining M2 and M3. We estimated concordance between M1 and the other methods using Cohen’s kappa (K) statistic. Results: Prevalence of the medical disorders based on M1 were lower than for either M2 or M3, depending on the disorder, and consistentlly lower for M4. For M1 and M4, moderate concordance (K ≥ 0.50) was observed only for psychiatric disorders (K = 0.52) and prior NAION (K=0.67). The prevalence and concordance for M1 and M4 for anemia, hypertension, diabetes and psychiatric disease were the only disorders that differed between females and males. For all methods, the prevalence varied widely across countries. Concordance for M1 and M4 varied and moderate concordance occurred for psychiatric disorders and prior NAION. Conclusion: Even with prospective, rigorously collected data, medical histories do not reliably identify all medical disorders. Adding the results of physical examination, laboratory tests, and medications increases the accuracy of reporting. This strategy could be adapted for clinical trials and electronic medical record disease-prevalence data mining.
- 11 April 2025
- Published by Elsevier in American Journal of Ophthalmology
- Vol. 276, 99-108
Deep Learning Approach Readily Differentiates Papilledema, Non-Arteritic Anterior Ischemic Optic Neuropathy, and Healthy Eyes
- 11 April 2025
- Published by Elsevier in American Journal of Ophthalmology
- Vol. 276, 99-108
Deep Learning Approach Readily Differentiates Papilledema, Non-Arteritic Anterior Ischemic Optic Neuropathy, and Healthy Eyes