Dmitri Volfson
Research Subjects
Dmitri Volfson has a total of 223 co-authors. Here are the top co-authors by number of shared publications:
- Lev S. Tsimring
Synthetic Biology Institute, University of California, San Diego
17 shared publications - Brian Tracey
Takeda Development Center Americas, Inc. , Cambridge, MA
7 shared publications Igor S. Aranson
Materials Science Division, Argonne National Laboratory, 9700 South Cass Avenue, Argonne, Illinois 60439, USA
5 shared publications- Marta Karas
Takeda Development Center Americas, Inc. , Cambridge, MA
5 shared publications - Francesco Onorati
Takeda Development Center Americas, Inc. , Cambridge, MA
4 shared publications Jeff Hasty
Department of Bioengineering and
4 shared publications
Latest Publications
- Open Access
- 18 August 2026
- Published by Oxford University Press (OUP) in SLEEP
An actigraphy-based algorithm to assess daytime napping in people with narcolepsy type 1
Study Objectives: Excessive daytime sleepiness is a defining symptom of narcolepsy, affecting daily functioning and quality of life. We developed and evaluated an actigraphy-based nap detection algorithm to objectively estimate daytime naps, a key manifestation of excessive daytime sleepiness. Materials and Methods: The nap detection algorithm was developed using data from the Multi-Ethnic Study of Atherosclerosis. This dataset included manually annotated naps from 2237 participants with 7 days of wrist-worn actigraphy. The nap detection algorithm was then applied in studies of participants with narcolepsy type 1 (NT1). Results: Our final nap detection algorithm yielded high specificity (Tau B: 93.2%) and accuracy (F1 Area: 83.9%). Applying our algorithm in an observational study (NCT04445129), participants with NT1 experienced 12.8 fewer nap-free days (p < .001) over 28 days and slept 34 minutes more (p < .001) during the day than controls. In a randomized clinical trial (NCT05687903), our algorithm demonstrated that participants with NT1 treated with an orexin receptor 2-selective agonist, oveporexton (TAK-861), experienced 6.1−11.9 additional nap-free days and slept 12−33 minutes fewer versus baseline. Conclusions: Overall, our nap detection algorithm retained sensitivity while minimizing false positives. It reliably identified nap-based phenotypes that differentiate individuals with NT1 from age and sex-matched healthy controls, and it demonstrated that an orexin receptor 2-selective agonist reduces napping in participants with NT1 to normative levels near those of healthy controls.
- 18 August 2026
- Published by Oxford University Press (OUP) in SLEEP
Open AccessAn actigraphy-based algorithm to assess daytime napping in people with narcolepsy type 1
Study Objectives: Excessive daytime sleepiness is a defining symptom of narcolepsy, affecting daily functioning and quality of life. We developed and evaluated an actigraphy-based nap detection algorithm to objectively estimate daytime naps, a key manifestation of excessive daytime sleepiness. Materials and Methods: The nap detection algorithm was developed using data from the Multi-Ethnic Study of Atherosclerosis. This dataset included manually annotated naps from 2237 participants with 7 days of wrist-worn actigraphy. The nap detection algorithm was then applied in studies of participants with narcolepsy type 1 (NT1). Results: Our final nap detection algorithm yielded high specificity (Tau B: 93.2%) and accuracy (F1 Area: 83.9%). Applying our algorithm in an observational study (NCT04445129), participants with NT1 experienced 12.8 fewer nap-free days (p < .001) over 28 days and slept 34 minutes more (p < .001) during the day than controls. In a randomized clinical trial (NCT05687903), our algorithm demonstrated that participants with NT1 treated with an orexin receptor 2-selective agonist, oveporexton (TAK-861), experienced 6.1−11.9 additional nap-free days and slept 12−33 minutes fewer versus baseline. Conclusions: Overall, our nap detection algorithm retained sensitivity while minimizing false positives. It reliably identified nap-based phenotypes that differentiate individuals with NT1 from age and sex-matched healthy controls, and it demonstrated that an orexin receptor 2-selective agonist reduces napping in participants with NT1 to normative levels near those of healthy controls.
- Open Access
- 7 July 2026
- Published by Oxford University Press (OUP) in SLEEP
- Vol. 49 (10)
Nocturnal Features and Daytime Characteristics in Narcolepsy: Reliability and Diagnostic Relevance for NT1 vs NT2
Study Objectives: Assess test–retest reliability of nighttime sleep and daytime Maintenance of Wakefulness Tests (MWTs) over a short time interval in drug-free individuals with narcolepsy; evaluate the diagnostic utility of features with good reliability. Methods: We analyzed the placebo arms from clinical trials in narcolepsy type 1 (NT1; n = 17; NCT05687903) and narcolepsy type 2 (NT2; n = 19; NCT05687916). Participants visited the clinic three times, four weeks apart, which included nocturnal polysomnography (nPSG) and MWT the following day. From the nPSG and daytime MWT, we derived 440 features and assessed test–retest reliability, visit effects, and NT1-NT2 differences. Results: Quantitative electroencephalography (qEEG) showed high reliability, with the top ten features showing reliability >0.88; the ten most reliable hypnogram features ranged from 0.68–0.78. qEEG and MWT sleep onset latencies (SOLs) were less reliable in NT2. N1 percentage, stage shift index, and wake after sleep onset were reliable and differed between diagnoses (r = 0.54–0.75; higher in NT1, p < 0.05). MWT SOLs at 2 p.m. and 4 p.m. demonstrated good reliability in NT1 (r = 0.89, 0.87) and differed between diagnoses (shorter in NT1; p = 0.037, 0.010). Hypnodensity Wake–REM mixtures were reliable (r = 0.83) and differed between diagnoses (higher in NT1, p = 0.001). Conclusions: qEEG features showed the strongest reliability but rarely distinguished diagnoses. Measures of sleep fragmentation and wakefulness reliably differentiated NT1 from NT2 and are practical for clinical assessments and trials. Hypnodensity features, though not yet used clinically, are reliable and may offer diagnostic value. NT2 exhibited greater variability in nighttime sleep and daytime wakefulness across visits.
- 7 July 2026
- Published by Oxford University Press (OUP) in SLEEP
- Vol. 49 (10)
Open AccessNocturnal Features and Daytime Characteristics in Narcolepsy: Reliability and Diagnostic Relevance for NT1 vs NT2
Study Objectives: Assess test–retest reliability of nighttime sleep and daytime Maintenance of Wakefulness Tests (MWTs) over a short time interval in drug-free individuals with narcolepsy; evaluate the diagnostic utility of features with good reliability. Methods: We analyzed the placebo arms from clinical trials in narcolepsy type 1 (NT1; n = 17; NCT05687903) and narcolepsy type 2 (NT2; n = 19; NCT05687916). Participants visited the clinic three times, four weeks apart, which included nocturnal polysomnography (nPSG) and MWT the following day. From the nPSG and daytime MWT, we derived 440 features and assessed test–retest reliability, visit effects, and NT1-NT2 differences. Results: Quantitative electroencephalography (qEEG) showed high reliability, with the top ten features showing reliability >0.88; the ten most reliable hypnogram features ranged from 0.68–0.78. qEEG and MWT sleep onset latencies (SOLs) were less reliable in NT2. N1 percentage, stage shift index, and wake after sleep onset were reliable and differed between diagnoses (r = 0.54–0.75; higher in NT1, p < 0.05). MWT SOLs at 2 p.m. and 4 p.m. demonstrated good reliability in NT1 (r = 0.89, 0.87) and differed between diagnoses (shorter in NT1; p = 0.037, 0.010). Hypnodensity Wake–REM mixtures were reliable (r = 0.83) and differed between diagnoses (higher in NT1, p = 0.001). Conclusions: qEEG features showed the strongest reliability but rarely distinguished diagnoses. Measures of sleep fragmentation and wakefulness reliably differentiated NT1 from NT2 and are practical for clinical assessments and trials. Hypnodensity features, though not yet used clinically, are reliable and may offer diagnostic value. NT2 exhibited greater variability in nighttime sleep and daytime wakefulness across visits.
- Open Access
Local REM sleep-N1-wake sleep stage mixing in narcolepsy type 1
Type 1 narcolepsy (NT1), a disorder caused by the loss of hypocretin/orexin transmission, is characterized by daytime sleepiness and symptoms where Rapid Eye Movement (REM) sleep, a state normally occurring from middle to late in the night, can intermingle with wakefulness. This results in cataplexy and sleep paralysis, episodes of muscle paralysis when awake, or in the generation of dream-like hallucinations and vivid dreaming, periods of visual imagery or sensory experiences that occur while awake, notably when falling asleep (hypnagogic hallucinations) or lingering dreams with over-realistic recall. Using deep learning of nocturnal sleep polysomnography (PSG) signals (EEG, EMG and EOG) applied to sleep stage scoring, we found that NT1 shows abnormally short wake to REM sleep transitions and occurrences of abnormal sleep stages probabilities of wake, REM sleep and N1 (very light NREM) sleep abnormally co-occurs (sleep stage mixing). Interestingly, although presence of these during sleep enables NT1 diagnosis with performances similar to gold standard diagnostic procedure, the multiple sleep latency test (MSLT), the cortical localization of these dissociations remains unclear. In this work, we used electrode specific predictions of sleep stages to explore if these are global or observed at the local cortical level. Surprisingly, although sleep stage mixing was preeminent between REM sleep, N1 and wake across all electrodes, it was found to fluctuate across locations, with stronger fluctuations found in frontal and central locations, notably in the dominant (left) hemisphere. The strongest single discriminator for NT1 was N1-REM stage mixing across central electrodes (C3-C4), showing 4.3-fold higher dissociation in NT1 patients (Cohen’s d = 0.61). Analysis of sleep stage dissociations across varying time scales revealed that windows lasting several minutes were most predictive of NT1 status, aligning with the duration of clinically reported symptoms of dissociated REM sleep in narcolepsy. Local N1-W-REM sleep dissociations correlated with CSF orexin/hypocretin levels and severity as measured using MSLT. The predominance of stage mixing in frontal and central regions, areas typically associated with executive and motor control, may contribute to the partial preservation of awareness during dissociated REM phenomena. Further, self-reports of hypnagogic hallucinations correlated best with dissociations involving occipital locations, in agreement with its usual visual content. Coherence analysis was also conducted but did not reveal additional insight. These results suggest that orexin deficiency destabilizes REM sleep organization across cortical projection area contributing both to REM sleep dissociation and to abnormal state transitions observed in NT1.
Open AccessLocal REM sleep-N1-wake sleep stage mixing in narcolepsy type 1
Type 1 narcolepsy (NT1), a disorder caused by the loss of hypocretin/orexin transmission, is characterized by daytime sleepiness and symptoms where Rapid Eye Movement (REM) sleep, a state normally occurring from middle to late in the night, can intermingle with wakefulness. This results in cataplexy and sleep paralysis, episodes of muscle paralysis when awake, or in the generation of dream-like hallucinations and vivid dreaming, periods of visual imagery or sensory experiences that occur while awake, notably when falling asleep (hypnagogic hallucinations) or lingering dreams with over-realistic recall. Using deep learning of nocturnal sleep polysomnography (PSG) signals (EEG, EMG and EOG) applied to sleep stage scoring, we found that NT1 shows abnormally short wake to REM sleep transitions and occurrences of abnormal sleep stages probabilities of wake, REM sleep and N1 (very light NREM) sleep abnormally co-occurs (sleep stage mixing). Interestingly, although presence of these during sleep enables NT1 diagnosis with performances similar to gold standard diagnostic procedure, the multiple sleep latency test (MSLT), the cortical localization of these dissociations remains unclear. In this work, we used electrode specific predictions of sleep stages to explore if these are global or observed at the local cortical level. Surprisingly, although sleep stage mixing was preeminent between REM sleep, N1 and wake across all electrodes, it was found to fluctuate across locations, with stronger fluctuations found in frontal and central locations, notably in the dominant (left) hemisphere. The strongest single discriminator for NT1 was N1-REM stage mixing across central electrodes (C3-C4), showing 4.3-fold higher dissociation in NT1 patients (Cohen’s d = 0.61). Analysis of sleep stage dissociations across varying time scales revealed that windows lasting several minutes were most predictive of NT1 status, aligning with the duration of clinically reported symptoms of dissociated REM sleep in narcolepsy. Local N1-W-REM sleep dissociations correlated with CSF orexin/hypocretin levels and severity as measured using MSLT. The predominance of stage mixing in frontal and central regions, areas typically associated with executive and motor control, may contribute to the partial preservation of awareness during dissociated REM phenomena. Further, self-reports of hypnagogic hallucinations correlated best with dissociations involving occipital locations, in agreement with its usual visual content. Coherence analysis was also conducted but did not reveal additional insight. These results suggest that orexin deficiency destabilizes REM sleep organization across cortical projection area contributing both to REM sleep dissociation and to abnormal state transitions observed in NT1.
- Open Access
- 1 January 2026
- Published by S. Karger AG in Digital Biomarkers
- Vol. 10 (1) , 113-124
Methods for Wearable-Derived Pulse Rate Measures and Their Application to Modeling the Relationship between Pulse Rate and Motor Activity in Narcolepsy Type 1
Introduction: Narcolepsy type 1 (NT1) is a chronic, rare neurological disorder caused by loss of orexin neurons that disrupts the sleep-wake cycle. Current research suggests a link between orexin neuropeptide deficiency and heart rate regulation. Wrist-worn devices enable continuous monitoring of physiological data, including heart rate. We aim to effectively utilize wearables in clinical research by developing robust methodologies to address challenges including data artifacts and missing data. Methods: Empatica wrist-worn smartwatch data were collected for 2 weeks at baseline in the TAK-861-2001 randomized, phase 2 clinical trial (NCT05687903) from participants with NT1. Epoch-level (minute-level) pulse rate data were rigorously screened for artifacts. Daily pulse rate measures (24-h, daytime, and nighttime averages) were derived. Numerical experiments were conducted to evaluate the stability of daily measures as a function of the proportion of valid minute-level data within a given day. Mean pulse rate and motor activity daily measures and the association between the two were quantified using linear mixed-effects models. Results: Participants (n = 110) had a mean age of 34.0 years and 51.8% were female. Approximately 23.5% of the epoch-level pulse rate observations were invalid due to non-wear, missing data, or artifacts. With a ≥70% threshold for valid daily wearable data, participants contributed an average of 10.3 days of data. Mean pulse rates were 72.8, 83.3, and 66.2 beats per minute and mean motor activities were 40.6, 61.5, and 10.3 epoch-level activity counts for 24-h, daytime, and nighttime averages, respectively. Pulse rate was significantly positively associated with motor activity daily measures, with up to 80% of pulse rate variance explained by motor activity levels and individual differences in baseline pulse rate and pulse rate-motor activity association. Conclusion: This work proposes a robust framework for preprocessing smartwatch pulse rate data for analyses in clinical research. The findings from our smartwatch data demonstrate a significant positive association between pulse rate and motor activity in participants with NT1 during a 2-week baseline period.
- 1 January 2026
- Published by S. Karger AG in Digital Biomarkers
- Vol. 10 (1) , 113-124
Open AccessMethods for Wearable-Derived Pulse Rate Measures and Their Application to Modeling the Relationship between Pulse Rate and Motor Activity in Narcolepsy Type 1
Introduction: Narcolepsy type 1 (NT1) is a chronic, rare neurological disorder caused by loss of orexin neurons that disrupts the sleep-wake cycle. Current research suggests a link between orexin neuropeptide deficiency and heart rate regulation. Wrist-worn devices enable continuous monitoring of physiological data, including heart rate. We aim to effectively utilize wearables in clinical research by developing robust methodologies to address challenges including data artifacts and missing data. Methods: Empatica wrist-worn smartwatch data were collected for 2 weeks at baseline in the TAK-861-2001 randomized, phase 2 clinical trial (NCT05687903) from participants with NT1. Epoch-level (minute-level) pulse rate data were rigorously screened for artifacts. Daily pulse rate measures (24-h, daytime, and nighttime averages) were derived. Numerical experiments were conducted to evaluate the stability of daily measures as a function of the proportion of valid minute-level data within a given day. Mean pulse rate and motor activity daily measures and the association between the two were quantified using linear mixed-effects models. Results: Participants (n = 110) had a mean age of 34.0 years and 51.8% were female. Approximately 23.5% of the epoch-level pulse rate observations were invalid due to non-wear, missing data, or artifacts. With a ≥70% threshold for valid daily wearable data, participants contributed an average of 10.3 days of data. Mean pulse rates were 72.8, 83.3, and 66.2 beats per minute and mean motor activities were 40.6, 61.5, and 10.3 epoch-level activity counts for 24-h, daytime, and nighttime averages, respectively. Pulse rate was significantly positively associated with motor activity daily measures, with up to 80% of pulse rate variance explained by motor activity levels and individual differences in baseline pulse rate and pulse rate-motor activity association. Conclusion: This work proposes a robust framework for preprocessing smartwatch pulse rate data for analyses in clinical research. The findings from our smartwatch data demonstrate a significant positive association between pulse rate and motor activity in participants with NT1 during a 2-week baseline period.
- Open Access
- 1 January 2026
- Published by S. Karger AG in Digital Biomarkers
- Vol. 10 (1) , 79-89
Functional Regression Methods for Estimating 24-Hour Ambulatory Blood Pressure and Heart Rate Parameters in Narcolepsy Type 1
Introduction: The nocturnal dip, a physiological drop in nocturnal blood pressure (BP), is driven by the autonomic nervous system. A reduction of <10% during nocturnal sleep versus daytime wakefulness is considered a “non-dipping” BP pattern and associated with increased cardiovascular disease risk in the general population. This study aimed to compare different methods for estimating BP and heart rate (HR) nocturnal dip from ambulatory BP monitoring (ABPM) data in individuals with narcolepsy type 1 (NT1). Methods: Baseline ABPM data were from participants with NT1 in the randomized TAK-994 phase 2 clinical trial (NCT04096560). Sleep period time (SPT) windows were estimated from raw accelerometer data overlapping with baseline and week 3 ABPM visits. Three approaches estimated BP and HR dip: (1) fixed-window, with daytime defined as 06:00 to 22:00, nighttime as 00:00 to 06:00, and dip defined as a drop from the daytime to nighttime window average; (2) 24-h pattern employing a two-component cosinor model to estimate a continuous 24-h pattern of BP and HR, and defining dip as a drop from pattern average to its lowest point; and (3) actigraphy-based, with dip defined as a drop from non-SPT to SPT average of BP and HR, utilizing algorithmically identified SPT aiming to best reflect participants’ actual sleep periods. Results: The analytic sample consisted of 31 participants with NT1. Comparing actigraphy-based dip with fixed-window and 24-h pattern dips, the 24-h pattern dip had higher Pearson’s correlation than the fixed-window dip across all three parameters (0.91 vs. 0.87, 0.88 vs. 0.68, and 0.88 vs. 0.56 for systolic BP [SBP], diastolic BP [DBP], and HR, respectively). We found substantial between- and within-participant variability in SPT timing and duration. A total of 61% of participants had a fixed-window SBP dip <10%, and 41% had a fixed-window DBP dip <10%. The 30th percentile of SBP/DBP dip varied substantially across calculation methods: 3.8%/8.6% (fixed-window), 6.8%/14.1% (24-h pattern), and 6.7%/12.1% (actigraphy-based). Conclusion: Estimated dip values from the 24-h pattern approach with a two-component cosinor model for BP and HR were strongly correlated with actigraphy-based dip values, which utilized an objective algorithm to identify participants’ sleep. The 24-h pattern approach offers a robust alternative to the fixed-window method for assessing dipping, especially in populations with sleep timing variations and disturbances, like NT1, and does not require simultaneous actigraphy measurement. The classification of a “non-dipper” varies depending on both the dip type (SBP vs. DBP) and the dip estimation method.
- 1 January 2026
- Published by S. Karger AG in Digital Biomarkers
- Vol. 10 (1) , 79-89
Open AccessFunctional Regression Methods for Estimating 24-Hour Ambulatory Blood Pressure and Heart Rate Parameters in Narcolepsy Type 1
Introduction: The nocturnal dip, a physiological drop in nocturnal blood pressure (BP), is driven by the autonomic nervous system. A reduction of <10% during nocturnal sleep versus daytime wakefulness is considered a “non-dipping” BP pattern and associated with increased cardiovascular disease risk in the general population. This study aimed to compare different methods for estimating BP and heart rate (HR) nocturnal dip from ambulatory BP monitoring (ABPM) data in individuals with narcolepsy type 1 (NT1). Methods: Baseline ABPM data were from participants with NT1 in the randomized TAK-994 phase 2 clinical trial (NCT04096560). Sleep period time (SPT) windows were estimated from raw accelerometer data overlapping with baseline and week 3 ABPM visits. Three approaches estimated BP and HR dip: (1) fixed-window, with daytime defined as 06:00 to 22:00, nighttime as 00:00 to 06:00, and dip defined as a drop from the daytime to nighttime window average; (2) 24-h pattern employing a two-component cosinor model to estimate a continuous 24-h pattern of BP and HR, and defining dip as a drop from pattern average to its lowest point; and (3) actigraphy-based, with dip defined as a drop from non-SPT to SPT average of BP and HR, utilizing algorithmically identified SPT aiming to best reflect participants’ actual sleep periods. Results: The analytic sample consisted of 31 participants with NT1. Comparing actigraphy-based dip with fixed-window and 24-h pattern dips, the 24-h pattern dip had higher Pearson’s correlation than the fixed-window dip across all three parameters (0.91 vs. 0.87, 0.88 vs. 0.68, and 0.88 vs. 0.56 for systolic BP [SBP], diastolic BP [DBP], and HR, respectively). We found substantial between- and within-participant variability in SPT timing and duration. A total of 61% of participants had a fixed-window SBP dip <10%, and 41% had a fixed-window DBP dip <10%. The 30th percentile of SBP/DBP dip varied substantially across calculation methods: 3.8%/8.6% (fixed-window), 6.8%/14.1% (24-h pattern), and 6.7%/12.1% (actigraphy-based). Conclusion: Estimated dip values from the 24-h pattern approach with a two-component cosinor model for BP and HR were strongly correlated with actigraphy-based dip values, which utilized an objective algorithm to identify participants’ sleep. The 24-h pattern approach offers a robust alternative to the fixed-window method for assessing dipping, especially in populations with sleep timing variations and disturbances, like NT1, and does not require simultaneous actigraphy measurement. The classification of a “non-dipper” varies depending on both the dip type (SBP vs. DBP) and the dip estimation method.
- Open Access
- 4 December 2025
- Published by Oxford University Press (OUP) in SLEEP
- Vol. 49 (3)
Classification and clustering on nocturnal polysomnography: distinctions and overlaps between central disorders of hypersomnolence
Study Objectives: Differential diagnosis of narcolepsy type 2 (NT2) from type 1 (NT1) and idiopathic hypersomnia (IH) is challenging due to overlapping symptoms. We developed an automated method using nocturnal polysomnography (nPSG) data to differentiate these conditions and clinical controls (CCs), and explored varying sleep phenotypes within NT1, NT2, IH, and CCs. Methods: We analyzed nPSG data from drug-free individuals with NT1, NT2, and IH, or CCs. Sleep features were derived at whole-night and per-quarter-night levels, including hypnogram, transition probability, hypnodensity, spindle, and quantitative electroencephalogram (qEEG) features. Random forest machine learning models were used for three classification tasks. Within-diagnosis clustering identified potential diagnosis subgroups. Results: The sample included 350 individuals (52% females; median age 30 years; 114 NT1, 90 NT2, 105 IH, and 41 CCs). Our models achieved area under the receiver operating characteristic curve values of 0.87, 0.79, and 0.82 for distinguishing NT2 from CCs, NT2 from IH, and IH from CCs, with corresponding F1 scores of 0.74, 0.71, and 0.69, respectively. qEEG features substantially contributed to model performance, distinguishing NT2 from IH. Cluster analysis revealed two NT1 subgroups (one showing more severe sleep disturbances), two NT2 subgroups (one trended toward NT1, the other toward IH), and two IH subgroups with differences in hypnodensity, qEEG, and spindle characteristics. Conclusions: Our exploratory findings demonstrate strong diagnosis classification performance from nPSG data alone, more easily distinguishing NT2 from CCs than from IH, and IH from CCs. The distinct NT2 subgroups suggest heterogeneity within NT2; further research is warranted to explore these patterns. Statement of Significance Accurate diagnoses of narcolepsy types 2 (NT2) and 1 and idiopathic hypersomnia (IH) remain challenging due to overlapping symptoms. We developed a machine learning model using drug-free nocturnal polysomnography data to automatically differentiate NT2 from clinical controls, NT2 from IH, and IH from clinical controls, with high accuracy. Our model leverages a rich set of sleep features, including spindle and quantitative electroencephalogram (qEEG) metrics. Furthermore, our analysis revealed distinct sleep phenotypes within each diagnosis, suggesting subtypes with varying levels of sleep disturbance and differences at qEEG and spindle levels. These findings provide a novel approach to classifying central disorders of hypersomnolence and suggest disease heterogeneity, which could lead to more accurate and timely diagnoses and personalized treatment strategies.
- 4 December 2025
- Published by Oxford University Press (OUP) in SLEEP
- Vol. 49 (3)
Open AccessClassification and clustering on nocturnal polysomnography: distinctions and overlaps between central disorders of hypersomnolence
Study Objectives: Differential diagnosis of narcolepsy type 2 (NT2) from type 1 (NT1) and idiopathic hypersomnia (IH) is challenging due to overlapping symptoms. We developed an automated method using nocturnal polysomnography (nPSG) data to differentiate these conditions and clinical controls (CCs), and explored varying sleep phenotypes within NT1, NT2, IH, and CCs. Methods: We analyzed nPSG data from drug-free individuals with NT1, NT2, and IH, or CCs. Sleep features were derived at whole-night and per-quarter-night levels, including hypnogram, transition probability, hypnodensity, spindle, and quantitative electroencephalogram (qEEG) features. Random forest machine learning models were used for three classification tasks. Within-diagnosis clustering identified potential diagnosis subgroups. Results: The sample included 350 individuals (52% females; median age 30 years; 114 NT1, 90 NT2, 105 IH, and 41 CCs). Our models achieved area under the receiver operating characteristic curve values of 0.87, 0.79, and 0.82 for distinguishing NT2 from CCs, NT2 from IH, and IH from CCs, with corresponding F1 scores of 0.74, 0.71, and 0.69, respectively. qEEG features substantially contributed to model performance, distinguishing NT2 from IH. Cluster analysis revealed two NT1 subgroups (one showing more severe sleep disturbances), two NT2 subgroups (one trended toward NT1, the other toward IH), and two IH subgroups with differences in hypnodensity, qEEG, and spindle characteristics. Conclusions: Our exploratory findings demonstrate strong diagnosis classification performance from nPSG data alone, more easily distinguishing NT2 from CCs than from IH, and IH from CCs. The distinct NT2 subgroups suggest heterogeneity within NT2; further research is warranted to explore these patterns. Statement of Significance Accurate diagnoses of narcolepsy types 2 (NT2) and 1 and idiopathic hypersomnia (IH) remain challenging due to overlapping symptoms. We developed a machine learning model using drug-free nocturnal polysomnography data to automatically differentiate NT2 from clinical controls, NT2 from IH, and IH from clinical controls, with high accuracy. Our model leverages a rich set of sleep features, including spindle and quantitative electroencephalogram (qEEG) metrics. Furthermore, our analysis revealed distinct sleep phenotypes within each diagnosis, suggesting subtypes with varying levels of sleep disturbance and differences at qEEG and spindle levels. These findings provide a novel approach to classifying central disorders of hypersomnolence and suggest disease heterogeneity, which could lead to more accurate and timely diagnoses and personalized treatment strategies.
- Open Access
- 2 July 2024
- Published by Oxford University Press (OUP) in SLEEP
- Vol. 47 (12)
Novel biomarkers derived from the Maintenance of Wakefulness Test as predictors of sleepiness and response to treatment
The Maintenance of Wakefulness Test (MWT) is a widely accepted objective test used to evaluate daytime somnolence and is commonly used in clinical studies evaluating novel therapeutics for excessive daytime sleepiness. In the latter, sleep onset latency (SOL) is typically the sole MWT endpoint. Here, we explored microsleeps, sleep probability measures derived from automated sleep scoring, and quantitative electroencephalography (qEEG) features as additional MWT biomarkers of daytime sleepiness, using data from a phase 1B trial of the selective orexin receptor 2 agonist danavorexton (TAK-925) in people with narcolepsy type 1 (NT1) or type 2 (NT2). Danavorexton treatment reduced the rate and duration of microsleeps during the MWT in NT1 (days 1 and 7; p ≤ 0.005) and microsleep rate in NT2 (days 1 and 7; p < 0.0001). Use of an EEG-sleep-staging−derived measure to determine the probability of wakefulness for each minute revealed a novel metric to track changes in daytime sleepiness, which were consistent with the θ/α ratio, a known biomarker of drowsiness. The slopes of line-fits to both the log-transformed sleepiness score or log-transformed θ/α ratio correlated well to (inverse) MWT SOL for NT1 (R = 0.93 and R = 0.83, respectively) and NT2 (R = 0.97 and R = 0.84, respectively), suggesting that individuals with narcolepsy have increased sleepiness immediately after lights-off. These analyses demonstrate that novel EEG-based biomarkers can augment SOL as predictors of sleepiness and its response to treatment and provide a novel framework for the analysis of wake EEG in hypersomnia disorders.
- 2 July 2024
- Published by Oxford University Press (OUP) in SLEEP
- Vol. 47 (12)
Open AccessNovel biomarkers derived from the Maintenance of Wakefulness Test as predictors of sleepiness and response to treatment
The Maintenance of Wakefulness Test (MWT) is a widely accepted objective test used to evaluate daytime somnolence and is commonly used in clinical studies evaluating novel therapeutics for excessive daytime sleepiness. In the latter, sleep onset latency (SOL) is typically the sole MWT endpoint. Here, we explored microsleeps, sleep probability measures derived from automated sleep scoring, and quantitative electroencephalography (qEEG) features as additional MWT biomarkers of daytime sleepiness, using data from a phase 1B trial of the selective orexin receptor 2 agonist danavorexton (TAK-925) in people with narcolepsy type 1 (NT1) or type 2 (NT2). Danavorexton treatment reduced the rate and duration of microsleeps during the MWT in NT1 (days 1 and 7; p ≤ 0.005) and microsleep rate in NT2 (days 1 and 7; p < 0.0001). Use of an EEG-sleep-staging−derived measure to determine the probability of wakefulness for each minute revealed a novel metric to track changes in daytime sleepiness, which were consistent with the θ/α ratio, a known biomarker of drowsiness. The slopes of line-fits to both the log-transformed sleepiness score or log-transformed θ/α ratio correlated well to (inverse) MWT SOL for NT1 (R = 0.93 and R = 0.83, respectively) and NT2 (R = 0.97 and R = 0.84, respectively), suggesting that individuals with narcolepsy have increased sleepiness immediately after lights-off. These analyses demonstrate that novel EEG-based biomarkers can augment SOL as predictors of sleepiness and its response to treatment and provide a novel framework for the analysis of wake EEG in hypersomnia disorders.
- Open Access
- 12 June 2024
- Published by Springer Nature in npj Parkinson's Disease
- Vol. 10 (1) , 1-11
Using a smartwatch and smartphone to assess early Parkinson’s disease in the WATCH-PD study over 12 months
Digital measures may provide objective, sensitive, real-world measures of disease progression in Parkinson’s disease (PD). However, multicenter longitudinal assessments of such measures are few. We recently demonstrated that baseline assessments of gait, tremor, finger tapping, and speech from a commercially available smartwatch, smartphone, and research-grade wearable sensors differed significantly between 82 individuals with early, untreated PD and 50 age-matched controls. Here, we evaluated the longitudinal change in these assessments over 12 months in a multicenter observational study using a generalized additive model, which permitted flexible modeling of at-home data. All measurements were included until participants started medications for PD. Over one year, individuals with early PD experienced significant declines in several measures of gait, an increase in the proportion of day with tremor, modest changes in speech, and few changes in psychomotor function. As measured by the smartwatch, the average (SD) arm swing in-clinic decreased from 25.9 (15.3) degrees at baseline to 19.9 degrees (13.7) at month 12 (P = 0.004). The proportion of awake time an individual with early PD had tremor increased from 19.3% (18.0%) to 25.6% (21.4%; P < 0.001). Activity, as measured by the number of steps taken per day, decreased from 3052 (1306) steps per day to 2331 (2010; P = 0.16), but this analysis was restricted to 10 participants due to the exclusion of those that had started PD medications and lost the data. The change of these digital measures over 12 months was generally larger than the corresponding change in individual items on the Movement Disorder Society—Unified Parkinson’s Disease Rating Scale but not greater than the change in the overall scale. Successful implementation of digital measures in future clinical trials will require improvements in study conduct, especially data capture. Nonetheless, gait and tremor measures derived from a commercially available smartwatch and smartphone hold promise for assessing the efficacy of therapeutics in early PD.
- 12 June 2024
- Published by Springer Nature in npj Parkinson's Disease
- Vol. 10 (1) , 1-11
Open AccessUsing a smartwatch and smartphone to assess early Parkinson’s disease in the WATCH-PD study over 12 months
Digital measures may provide objective, sensitive, real-world measures of disease progression in Parkinson’s disease (PD). However, multicenter longitudinal assessments of such measures are few. We recently demonstrated that baseline assessments of gait, tremor, finger tapping, and speech from a commercially available smartwatch, smartphone, and research-grade wearable sensors differed significantly between 82 individuals with early, untreated PD and 50 age-matched controls. Here, we evaluated the longitudinal change in these assessments over 12 months in a multicenter observational study using a generalized additive model, which permitted flexible modeling of at-home data. All measurements were included until participants started medications for PD. Over one year, individuals with early PD experienced significant declines in several measures of gait, an increase in the proportion of day with tremor, modest changes in speech, and few changes in psychomotor function. As measured by the smartwatch, the average (SD) arm swing in-clinic decreased from 25.9 (15.3) degrees at baseline to 19.9 degrees (13.7) at month 12 (P = 0.004). The proportion of awake time an individual with early PD had tremor increased from 19.3% (18.0%) to 25.6% (21.4%; P < 0.001). Activity, as measured by the number of steps taken per day, decreased from 3052 (1306) steps per day to 2331 (2010; P = 0.16), but this analysis was restricted to 10 participants due to the exclusion of those that had started PD medications and lost the data. The change of these digital measures over 12 months was generally larger than the corresponding change in individual items on the Movement Disorder Society—Unified Parkinson’s Disease Rating Scale but not greater than the change in the overall scale. Successful implementation of digital measures in future clinical trials will require improvements in study conduct, especially data capture. Nonetheless, gait and tremor measures derived from a commercially available smartwatch and smartphone hold promise for assessing the efficacy of therapeutics in early PD.
- Open Access
- 31 May 2024
- Published by Public Library of Science (PLoS) in PLOS ONE
- Vol. 19 (5) , e0304415
Cardiac manifestations of Fabry disease in G3Stg/GlaKO and GlaKO mouse models–Translation to Fabry disease patients
Fabry disease (FD) is an X-linked disorder of glycosphingolipid metabolism caused by mutations in the GLA gene encoding alpha-galactosidase A (α-Gal). Loss of α-Gal activity leads to progressive lysosomal accumulation of α-Gal substrate, predominately globotriaosylceramide (Gb3) and its deacylated derivative globotriaosylsphingosine (lyso‐Gb3). FD manifestations include early onset neuropathic pain, gastrointestinal symptoms, and later onset life-threatening renal, cardiovascular and cerebrovascular disorders. Current treatments can preserve kidney function but are not very effective in preventing progression of cardiovascular pathology which remains the most common cause of premature death in FD patients. There is a significant need for a translational model that could be used for testing cardiac efficacy of new drugs. Two mouse models of FD have been developed. The α-Gal A-knockout (GlaKO) model is characterized by progressive tissue accumulation of Gb3 and lyso-Gb3 but does not develop any Fabry pathology besides mild peripheral neuropathy. Reports of minor cardiac function abnormalities in GlaKO model are inconsistent between different studies. Recently, G3Stg/GlaKO was generated by crossbreeding GlaKO with transgenic mice expressing human Gb3 synthase. G3Stg/GlaKO demonstrate higher tissue substrate accumulation and develop cellular and tissue pathologies. Functional renal pathology analogous to that found in early stages of FD has also been described in this model. The objective of this study is to characterize cardiac phenotype in GlaKO and G3Stg/GlaKO mice using echocardiography. Longitudinal assessments of cardiac wall thickness, mass and function were performed in GlaKO and wild-type (WT) littermate controls from 5–13 months of age. G3Stg/GlaKO and WT mice were assessed between 27–28 weeks of age due to their shortened lifespan. Several cardiomyopathy characteristics of early Fabry pathology were found in GlaKO mice, including mild cardiomegaly [up-to-25% increase in left ventricular (LV mass)] with no significant LV wall thickening. The LV internal diameter was significantly wider (up-to-24% increase at 9-months), when compared to the age-matched WT. In addition, there were significant increases in the end-systolic, end-diastolic volumes and stroke volume, suggesting volume overload. Significant reduction in Global longitudinal strain (GLS) measuring local myofiber contractility of the LV was also detected at 13-months. Similar GLS reduction was also reported in FD patients. Parameters such as ejection fraction, fractional shortening and cardiac output were either only slightly affected or were not different from controls. On the other hand, some of the cardiac findings in G3Stg/GlaKO mice were inconsistent with Fabry cardiomyopathy seen in FD patients. This could be potentially an artifact of the Gb3 synthase overexpression under a strong ubiquitous promoter. In conclusion, GlaKO mouse model presents mild cardiomegaly, mild cardiac dysfunction, but significant cardiac volume overload and functional changes in GLS that can be used as translational biomarkers to determine cardiac efficacy of novel treatment modalities. The level of tissue Gb3 accumulation in G3Stg/GlaKO mouse more closely recapitulates the level of substrate accumulation in FD patients and may provide better translatability of the efficacy of new therapeutics in clearing pathological substrates from cardiac tissues. But interpretation of the effect of treatment on cardiac structure and function in this model should be approached with caution.
- 31 May 2024
- Published by Public Library of Science (PLoS) in PLOS ONE
- Vol. 19 (5) , e0304415
Open AccessCardiac manifestations of Fabry disease in G3Stg/GlaKO and GlaKO mouse models–Translation to Fabry disease patients
Fabry disease (FD) is an X-linked disorder of glycosphingolipid metabolism caused by mutations in the GLA gene encoding alpha-galactosidase A (α-Gal). Loss of α-Gal activity leads to progressive lysosomal accumulation of α-Gal substrate, predominately globotriaosylceramide (Gb3) and its deacylated derivative globotriaosylsphingosine (lyso‐Gb3). FD manifestations include early onset neuropathic pain, gastrointestinal symptoms, and later onset life-threatening renal, cardiovascular and cerebrovascular disorders. Current treatments can preserve kidney function but are not very effective in preventing progression of cardiovascular pathology which remains the most common cause of premature death in FD patients. There is a significant need for a translational model that could be used for testing cardiac efficacy of new drugs. Two mouse models of FD have been developed. The α-Gal A-knockout (GlaKO) model is characterized by progressive tissue accumulation of Gb3 and lyso-Gb3 but does not develop any Fabry pathology besides mild peripheral neuropathy. Reports of minor cardiac function abnormalities in GlaKO model are inconsistent between different studies. Recently, G3Stg/GlaKO was generated by crossbreeding GlaKO with transgenic mice expressing human Gb3 synthase. G3Stg/GlaKO demonstrate higher tissue substrate accumulation and develop cellular and tissue pathologies. Functional renal pathology analogous to that found in early stages of FD has also been described in this model. The objective of this study is to characterize cardiac phenotype in GlaKO and G3Stg/GlaKO mice using echocardiography. Longitudinal assessments of cardiac wall thickness, mass and function were performed in GlaKO and wild-type (WT) littermate controls from 5–13 months of age. G3Stg/GlaKO and WT mice were assessed between 27–28 weeks of age due to their shortened lifespan. Several cardiomyopathy characteristics of early Fabry pathology were found in GlaKO mice, including mild cardiomegaly [up-to-25% increase in left ventricular (LV mass)] with no significant LV wall thickening. The LV internal diameter was significantly wider (up-to-24% increase at 9-months), when compared to the age-matched WT. In addition, there were significant increases in the end-systolic, end-diastolic volumes and stroke volume, suggesting volume overload. Significant reduction in Global longitudinal strain (GLS) measuring local myofiber contractility of the LV was also detected at 13-months. Similar GLS reduction was also reported in FD patients. Parameters such as ejection fraction, fractional shortening and cardiac output were either only slightly affected or were not different from controls. On the other hand, some of the cardiac findings in G3Stg/GlaKO mice were inconsistent with Fabry cardiomyopathy seen in FD patients. This could be potentially an artifact of the Gb3 synthase overexpression under a strong ubiquitous promoter. In conclusion, GlaKO mouse model presents mild cardiomegaly, mild cardiac dysfunction, but significant cardiac volume overload and functional changes in GLS that can be used as translational biomarkers to determine cardiac efficacy of novel treatment modalities. The level of tissue Gb3 accumulation in G3Stg/GlaKO mouse more closely recapitulates the level of substrate accumulation in FD patients and may provide better translatability of the efficacy of new therapeutics in clearing pathological substrates from cardiac tissues. But interpretation of the effect of treatment on cardiac structure and function in this model should be approached with caution.
- Open Access
- 21 December 2021
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Biomedical and Health Informatics
- Vol. 26 (6) , 2787-2795
Voice Biomarkers of Recovery From Acute Respiratory Illness
Voice analysis is an emerging technology which has the potential to provide low-cost, at-home monitoring of symptoms associated with a variety of health conditions. While voice has received significant attention for monitoring neurological disease, few studies have focused on voice changes related to flu-like symptoms. Herein, we investigate the relationship between changes in acoustic features of voice and self-reported symptoms during recovery from a flu-like illness in a cohort of 29 subjects. Acoustic features were automatically extracted from “sick” and “well” visit data collected in the laboratory setting, and feature down-selection was used to identify those that change significantly between visits. The selected acoustic features were extracted from at-home data and used to construct a combined distance metric that correlated with self-reported symptoms (0.63 rank correlation). Changes in self-reported symptoms corresponding to 10% of the ordinal scale used in the study were detected with an area under the curve of 0.72. The results show that acoustic features derived from voice recordings may provide an objective measure for diagnosing and monitoring symptoms of respiratory illnesses.
- 21 December 2021
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Biomedical and Health Informatics
- Vol. 26 (6) , 2787-2795
Open AccessVoice Biomarkers of Recovery From Acute Respiratory Illness
Voice analysis is an emerging technology which has the potential to provide low-cost, at-home monitoring of symptoms associated with a variety of health conditions. While voice has received significant attention for monitoring neurological disease, few studies have focused on voice changes related to flu-like symptoms. Herein, we investigate the relationship between changes in acoustic features of voice and self-reported symptoms during recovery from a flu-like illness in a cohort of 29 subjects. Acoustic features were automatically extracted from “sick” and “well” visit data collected in the laboratory setting, and feature down-selection was used to identify those that change significantly between visits. The selected acoustic features were extracted from at-home data and used to construct a combined distance metric that correlated with self-reported symptoms (0.63 rank correlation). Changes in self-reported symptoms corresponding to 10% of the ordinal scale used in the study were detected with an area under the curve of 0.72. The results show that acoustic features derived from voice recordings may provide an objective measure for diagnosing and monitoring symptoms of respiratory illnesses.