Wrist-worn system detected convulsive seizures in home setting

 

HOUSTON – A wrist-worn device and smartphone-based alert system worn by a patient outside of the epilepsy monitoring unit for 3 months effectively detected convulsive seizure events and minimized false alarms, results from a long-term study found.

The automated device, known as Embrace, relies on an accelerometer and electrodermal activity to detect convulsive seizures. It pairs a wrist-worn device with a smartphone-based application that provides an alert to designated caregivers when an unusual event is detected. Embrace, which is being developed by Empatica Inc., received clearance by the European Union as a medical device for convulsive seizure detection but is still under consideration by the Food and Drug Administration. Previous reports evaluating the automated detector used by Embrace have always relied on data from epilepsy monitoring units (EMUs), where it achieved sensitivity scores ranging from 92% to 100% and 0.15-2.02 false alarms per day in detecting convulsive seizures.

Andy Ryan

“The EMU is expected to be an easier environment for obtaining better performance because people tend to be less active there, so it’s possible that EMU-based results reported in the past may be too optimistic compared to real life,” Rosalind W. Picard, ScD, said in an interview in advance of the annual meeting of the annual meeting of the American Epilepsy Society. “We wanted to see how it changed in real life.”

To find out, Dr. Picard, chief scientist at Empatica, and her associates conducted a case study of a 14-year-old patient with Dravet syndrome who was enrolled in a trial of Embrace in the outpatient setting. No data from this patient were used in training the system. The patient’s caregiver was asked to annotate the occurrence of each convulsive seizure and any activity that generated an alert. The number of false alerts was obtained by subtracting the number of correctly recognized convulsive seizures from the total alerts fired by the device. Sensitivity was the percentage of convulsive seizures that automatically triggered an alert.

Courtesy Dr. Rosalind W. Picard

The researchers found that, after 113 days of use, Embrace detected 22 of the patient’s 24 convulsive seizures, for a sensitivity of 92% and a false-alarm rate of 0.35 false alarms per day worn. Most days had no false alarms. Only 2 of the 113 days had more than two false alarms per day. “We found the performance in this long-term, home-use case study to be as good as our earlier reported results in the EMU,” said Dr. Picard, who is also professor of affective computing at the Massachusetts Institute of Technology Media Lab, Cambridge.

“Embrace can work for patients outside the EMU, detecting events and issuing alerts through a paired smartphone to a designated caregiver. Also, for clinicians who understand machine learning, we should be clear that we used the strongest test possible: none of the patient’s data were used to train the algorithm that was used, nor was it tuned in any way special for this patient. Had we done so, the results could have possibly appeared even better. However, we chose to use this test of generalization to better get a sense of how it might work on new patients, whose data had never been seen before.”

Dr. Picard acknowledged certain limitations of the study, including the fact that it was a single case and that the patient had Dravet syndrome. She disclosed that she is a cofounder of Empatica and owns shares in the company.

dbrunk@frontlinemedcom.com

 

HOUSTON – A wrist-worn device and smartphone-based alert system worn by a patient outside of the epilepsy monitoring unit for 3 months effectively detected convulsive seizure events and minimized false alarms, results from a long-term study found.

The automated device, known as Embrace, relies on an accelerometer and electrodermal activity to detect convulsive seizures. It pairs a wrist-worn device with a smartphone-based application that provides an alert to designated caregivers when an unusual event is detected. Embrace, which is being developed by Empatica Inc., received clearance by the European Union as a medical device for convulsive seizure detection but is still under consideration by the Food and Drug Administration. Previous reports evaluating the automated detector used by Embrace have always relied on data from epilepsy monitoring units (EMUs), where it achieved sensitivity scores ranging from 92% to 100% and 0.15-2.02 false alarms per day in detecting convulsive seizures.

Andy Ryan

“The EMU is expected to be an easier environment for obtaining better performance because people tend to be less active there, so it’s possible that EMU-based results reported in the past may be too optimistic compared to real life,” Rosalind W. Picard, ScD, said in an interview in advance of the annual meeting of the annual meeting of the American Epilepsy Society. “We wanted to see how it changed in real life.”

To find out, Dr. Picard, chief scientist at Empatica, and her associates conducted a case study of a 14-year-old patient with Dravet syndrome who was enrolled in a trial of Embrace in the outpatient setting. No data from this patient were used in training the system. The patient’s caregiver was asked to annotate the occurrence of each convulsive seizure and any activity that generated an alert. The number of false alerts was obtained by subtracting the number of correctly recognized convulsive seizures from the total alerts fired by the device. Sensitivity was the percentage of convulsive seizures that automatically triggered an alert.

Courtesy Dr. Rosalind W. Picard

The researchers found that, after 113 days of use, Embrace detected 22 of the patient’s 24 convulsive seizures, for a sensitivity of 92% and a false-alarm rate of 0.35 false alarms per day worn. Most days had no false alarms. Only 2 of the 113 days had more than two false alarms per day. “We found the performance in this long-term, home-use case study to be as good as our earlier reported results in the EMU,” said Dr. Picard, who is also professor of affective computing at the Massachusetts Institute of Technology Media Lab, Cambridge.

“Embrace can work for patients outside the EMU, detecting events and issuing alerts through a paired smartphone to a designated caregiver. Also, for clinicians who understand machine learning, we should be clear that we used the strongest test possible: none of the patient’s data were used to train the algorithm that was used, nor was it tuned in any way special for this patient. Had we done so, the results could have possibly appeared even better. However, we chose to use this test of generalization to better get a sense of how it might work on new patients, whose data had never been seen before.”

Dr. Picard acknowledged certain limitations of the study, including the fact that it was a single case and that the patient had Dravet syndrome. She disclosed that she is a cofounder of Empatica and owns shares in the company.

dbrunk@frontlinemedcom.com

 

HOUSTON – A wrist-worn device and smartphone-based alert system worn by a patient outside of the epilepsy monitoring unit for 3 months effectively detected convulsive seizure events and minimized false alarms, results from a long-term study found.

The automated device, known as Embrace, relies on an accelerometer and electrodermal activity to detect convulsive seizures. It pairs a wrist-worn device with a smartphone-based application that provides an alert to designated caregivers when an unusual event is detected. Embrace, which is being developed by Empatica Inc., received clearance by the European Union as a medical device for convulsive seizure detection but is still under consideration by the Food and Drug Administration. Previous reports evaluating the automated detector used by Embrace have always relied on data from epilepsy monitoring units (EMUs), where it achieved sensitivity scores ranging from 92% to 100% and 0.15-2.02 false alarms per day in detecting convulsive seizures.

Andy Ryan

“The EMU is expected to be an easier environment for obtaining better performance because people tend to be less active there, so it’s possible that EMU-based results reported in the past may be too optimistic compared to real life,” Rosalind W. Picard, ScD, said in an interview in advance of the annual meeting of the annual meeting of the American Epilepsy Society. “We wanted to see how it changed in real life.”

To find out, Dr. Picard, chief scientist at Empatica, and her associates conducted a case study of a 14-year-old patient with Dravet syndrome who was enrolled in a trial of Embrace in the outpatient setting. No data from this patient were used in training the system. The patient’s caregiver was asked to annotate the occurrence of each convulsive seizure and any activity that generated an alert. The number of false alerts was obtained by subtracting the number of correctly recognized convulsive seizures from the total alerts fired by the device. Sensitivity was the percentage of convulsive seizures that automatically triggered an alert.

Courtesy Dr. Rosalind W. Picard

The researchers found that, after 113 days of use, Embrace detected 22 of the patient’s 24 convulsive seizures, for a sensitivity of 92% and a false-alarm rate of 0.35 false alarms per day worn. Most days had no false alarms. Only 2 of the 113 days had more than two false alarms per day. “We found the performance in this long-term, home-use case study to be as good as our earlier reported results in the EMU,” said Dr. Picard, who is also professor of affective computing at the Massachusetts Institute of Technology Media Lab, Cambridge.

“Embrace can work for patients outside the EMU, detecting events and issuing alerts through a paired smartphone to a designated caregiver. Also, for clinicians who understand machine learning, we should be clear that we used the strongest test possible: none of the patient’s data were used to train the algorithm that was used, nor was it tuned in any way special for this patient. Had we done so, the results could have possibly appeared even better. However, we chose to use this test of generalization to better get a sense of how it might work on new patients, whose data had never been seen before.”

Dr. Picard acknowledged certain limitations of the study, including the fact that it was a single case and that the patient had Dravet syndrome. She disclosed that she is a cofounder of Empatica and owns shares in the company.

dbrunk@frontlinemedcom.com