Explainable AI for wearable seizure logging: Impact of data quality, patient age, and antiseizure medication on performance
Objective seizure count estimates are crucial for ambulatory epilepsy management. Wearables have shown promise for the detection of tonic-clonic seizures but may suffer from false alarms and undetected seizures. Seizure signatures recorded by wearables often occur over prolonged periods, including i...
Saved in:
| Published in: | Seizure (London, England) Vol. 110; p. 99 |
|---|---|
| Main Authors: | , , , , |
| Format: | Journal Article |
| Language: | English |
| Published: |
England
01.08.2023
|
| Subjects: | |
| ISSN: | 1532-2688, 1532-2688 |
| Online Access: | Get more information |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Abstract | Objective seizure count estimates are crucial for ambulatory epilepsy management. Wearables have shown promise for the detection of tonic-clonic seizures but may suffer from false alarms and undetected seizures. Seizure signatures recorded by wearables often occur over prolonged periods, including increased levels of electrodermal activity and heart rate long after seizure EEG onset, however, previous detection methods only partially exploited these signatures. Understanding the utility of these prolonged signatures for seizure count estimation and what factors generally determine seizure logging performance, including the role of data quality vs. algorithm performance, is thus crucial for improving wearables-based epilepsy monitoring and determining which patients benefit most from this technology.
In this retrospective study we examined 76 pediatric epilepsy patients during multiday video-EEG monitoring equipped with a wearable (Empatica E4; records of electrodermal activity, EDA, accelerometry, ACC, heart rate, HR; 1983 h total recording time; 45 tonic-clonic seizures). To log seizures on prolonged data trends, we applied deep learning on continuous overlapping 1-hour segments of multimodal data in a leave-one-subject-out approach. We systematically examined factors influencing logging performance, including patient age, antiseizure medication (ASM) load, seizure type and duration, and data artifacts. To gain insights into algorithm function and feature importance we applied Uniform Manifold Approximation and Projection (UMAP, to represent the separability of learned features) and SHapley Additive exPlanations (SHAP, to represent the most informative data signatures).
Performance for tonic-clonic seizure logging increased systematically with patient age (AUC 0.61 for patients 〈 11 years, AUC 0.77 for patients between 11-15 years, AUC 0.85 for patients 〉 15 years). Across all ages, AUC was 0.75 corresponding to a sensitivity of 0.52 and a false alarm rate of 0.28/24 h. Seizures under high ASM load or with shorter duration were detected worse (P=.025, P=.033, respectively). UMAP visualized discriminatory power at the individual patient level, SHAP analyses identified clonic motor activity and peri/postictal increases in HR and EDA as most informative. In contrast, in missed seizures, these features were absent indicating that recording quality but not the algorithm caused the low sensitivity in these patients.
Our results demonstrate the utility of prolonged, postictal data segments for seizure logging, contribute to algorithm explainability and point to influencing factors, including high ASM dose and short seizure duration. Collectively, these results may help to identify patients who particularly benefit from such technology. |
|---|---|
| AbstractList | Objective seizure count estimates are crucial for ambulatory epilepsy management. Wearables have shown promise for the detection of tonic-clonic seizures but may suffer from false alarms and undetected seizures. Seizure signatures recorded by wearables often occur over prolonged periods, including increased levels of electrodermal activity and heart rate long after seizure EEG onset, however, previous detection methods only partially exploited these signatures. Understanding the utility of these prolonged signatures for seizure count estimation and what factors generally determine seizure logging performance, including the role of data quality vs. algorithm performance, is thus crucial for improving wearables-based epilepsy monitoring and determining which patients benefit most from this technology.OBJECTIVEObjective seizure count estimates are crucial for ambulatory epilepsy management. Wearables have shown promise for the detection of tonic-clonic seizures but may suffer from false alarms and undetected seizures. Seizure signatures recorded by wearables often occur over prolonged periods, including increased levels of electrodermal activity and heart rate long after seizure EEG onset, however, previous detection methods only partially exploited these signatures. Understanding the utility of these prolonged signatures for seizure count estimation and what factors generally determine seizure logging performance, including the role of data quality vs. algorithm performance, is thus crucial for improving wearables-based epilepsy monitoring and determining which patients benefit most from this technology.In this retrospective study we examined 76 pediatric epilepsy patients during multiday video-EEG monitoring equipped with a wearable (Empatica E4; records of electrodermal activity, EDA, accelerometry, ACC, heart rate, HR; 1983 h total recording time; 45 tonic-clonic seizures). To log seizures on prolonged data trends, we applied deep learning on continuous overlapping 1-hour segments of multimodal data in a leave-one-subject-out approach. We systematically examined factors influencing logging performance, including patient age, antiseizure medication (ASM) load, seizure type and duration, and data artifacts. To gain insights into algorithm function and feature importance we applied Uniform Manifold Approximation and Projection (UMAP, to represent the separability of learned features) and SHapley Additive exPlanations (SHAP, to represent the most informative data signatures).METHODSIn this retrospective study we examined 76 pediatric epilepsy patients during multiday video-EEG monitoring equipped with a wearable (Empatica E4; records of electrodermal activity, EDA, accelerometry, ACC, heart rate, HR; 1983 h total recording time; 45 tonic-clonic seizures). To log seizures on prolonged data trends, we applied deep learning on continuous overlapping 1-hour segments of multimodal data in a leave-one-subject-out approach. We systematically examined factors influencing logging performance, including patient age, antiseizure medication (ASM) load, seizure type and duration, and data artifacts. To gain insights into algorithm function and feature importance we applied Uniform Manifold Approximation and Projection (UMAP, to represent the separability of learned features) and SHapley Additive exPlanations (SHAP, to represent the most informative data signatures).Performance for tonic-clonic seizure logging increased systematically with patient age (AUC 0.61 for patients 〈 11 years, AUC 0.77 for patients between 11-15 years, AUC 0.85 for patients 〉 15 years). Across all ages, AUC was 0.75 corresponding to a sensitivity of 0.52 and a false alarm rate of 0.28/24 h. Seizures under high ASM load or with shorter duration were detected worse (P=.025, P=.033, respectively). UMAP visualized discriminatory power at the individual patient level, SHAP analyses identified clonic motor activity and peri/postictal increases in HR and EDA as most informative. In contrast, in missed seizures, these features were absent indicating that recording quality but not the algorithm caused the low sensitivity in these patients.RESULTSPerformance for tonic-clonic seizure logging increased systematically with patient age (AUC 0.61 for patients 〈 11 years, AUC 0.77 for patients between 11-15 years, AUC 0.85 for patients 〉 15 years). Across all ages, AUC was 0.75 corresponding to a sensitivity of 0.52 and a false alarm rate of 0.28/24 h. Seizures under high ASM load or with shorter duration were detected worse (P=.025, P=.033, respectively). UMAP visualized discriminatory power at the individual patient level, SHAP analyses identified clonic motor activity and peri/postictal increases in HR and EDA as most informative. In contrast, in missed seizures, these features were absent indicating that recording quality but not the algorithm caused the low sensitivity in these patients.Our results demonstrate the utility of prolonged, postictal data segments for seizure logging, contribute to algorithm explainability and point to influencing factors, including high ASM dose and short seizure duration. Collectively, these results may help to identify patients who particularly benefit from such technology.SIGNIFICANCEOur results demonstrate the utility of prolonged, postictal data segments for seizure logging, contribute to algorithm explainability and point to influencing factors, including high ASM dose and short seizure duration. Collectively, these results may help to identify patients who particularly benefit from such technology. Objective seizure count estimates are crucial for ambulatory epilepsy management. Wearables have shown promise for the detection of tonic-clonic seizures but may suffer from false alarms and undetected seizures. Seizure signatures recorded by wearables often occur over prolonged periods, including increased levels of electrodermal activity and heart rate long after seizure EEG onset, however, previous detection methods only partially exploited these signatures. Understanding the utility of these prolonged signatures for seizure count estimation and what factors generally determine seizure logging performance, including the role of data quality vs. algorithm performance, is thus crucial for improving wearables-based epilepsy monitoring and determining which patients benefit most from this technology. In this retrospective study we examined 76 pediatric epilepsy patients during multiday video-EEG monitoring equipped with a wearable (Empatica E4; records of electrodermal activity, EDA, accelerometry, ACC, heart rate, HR; 1983 h total recording time; 45 tonic-clonic seizures). To log seizures on prolonged data trends, we applied deep learning on continuous overlapping 1-hour segments of multimodal data in a leave-one-subject-out approach. We systematically examined factors influencing logging performance, including patient age, antiseizure medication (ASM) load, seizure type and duration, and data artifacts. To gain insights into algorithm function and feature importance we applied Uniform Manifold Approximation and Projection (UMAP, to represent the separability of learned features) and SHapley Additive exPlanations (SHAP, to represent the most informative data signatures). Performance for tonic-clonic seizure logging increased systematically with patient age (AUC 0.61 for patients 〈 11 years, AUC 0.77 for patients between 11-15 years, AUC 0.85 for patients 〉 15 years). Across all ages, AUC was 0.75 corresponding to a sensitivity of 0.52 and a false alarm rate of 0.28/24 h. Seizures under high ASM load or with shorter duration were detected worse (P=.025, P=.033, respectively). UMAP visualized discriminatory power at the individual patient level, SHAP analyses identified clonic motor activity and peri/postictal increases in HR and EDA as most informative. In contrast, in missed seizures, these features were absent indicating that recording quality but not the algorithm caused the low sensitivity in these patients. Our results demonstrate the utility of prolonged, postictal data segments for seizure logging, contribute to algorithm explainability and point to influencing factors, including high ASM dose and short seizure duration. Collectively, these results may help to identify patients who particularly benefit from such technology. |
| Author | Jackson, Michele Meisel, Christian Halimeh, Mustafa Loddenkemper, Tobias Vieluf, Solveig |
| Author_xml | – sequence: 1 givenname: Mustafa surname: Halimeh fullname: Halimeh, Mustafa organization: Computational Neurology Lab, Department of Neurology, Charité - Universitätsmedizin Berlin, Charitéplatz 1, 10117 Berlin, Germany – sequence: 2 givenname: Michele surname: Jackson fullname: Jackson, Michele organization: Division of Epilepsy and Clinical Neurophysiology, Boston Children's Hospital, Harvard Medical School, 300 Longwood Ave, Boston, MA 02115, United States – sequence: 3 givenname: Solveig surname: Vieluf fullname: Vieluf, Solveig organization: Division of Epilepsy and Clinical Neurophysiology, Boston Children's Hospital, Harvard Medical School, 300 Longwood Ave, Boston, MA 02115, United States – sequence: 4 givenname: Tobias surname: Loddenkemper fullname: Loddenkemper, Tobias organization: Division of Epilepsy and Clinical Neurophysiology, Boston Children's Hospital, Harvard Medical School, 300 Longwood Ave, Boston, MA 02115, United States – sequence: 5 givenname: Christian surname: Meisel fullname: Meisel, Christian email: christian@meisel.de organization: Computational Neurology Lab, Department of Neurology, Charité - Universitätsmedizin Berlin, Charitéplatz 1, 10117 Berlin, Germany; Center for Stroke Research Berlin, Berlin, Germany; Berlin Institute of Health, Berlin, Germany; NeuroCure Cluster of Excellence, Charité - Universitätsmedizin Berlin, Berlin, Germany; Bernstein Center for Computational Neuroscience, Berlin, Germany. Electronic address: christian@meisel.de |
| BackLink | https://www.ncbi.nlm.nih.gov/pubmed/37336056$$D View this record in MEDLINE/PubMed |
| BookMark | eNpNkE1Lw0AQhhep2A_9CcoePTRxs5uPjbdSqhYKXvQcZjeTsCXZpMkGrf55g7YgzDAz8LzzDjMnE9tYJOQ2YH7Agvhh7_dovoYOfc648FnsM8YvyCyIBPd4LOXkXz8l877fM8bSMBBXZCoSIWIWxTPyvflsKzAWVIV0taVF09EPhO53PjnQqilLY8tHuq1b0I42Bc3BAT0MUBl3XNIWnEHrKJS4pGDzMZ05q2vMjR6BxtIxWuxGjxqsxmtyWUDV482pLsj70-Zt_eLtXp-369XO02GQOk9zlIWUQgmRMKlEkhaFisICEh1EOWe6YFKECnMdpjxUaaKiVGktImCKc5HzBbn_29t2zWHA3mW16TVWFVhshj7jkiejkgk2oncndFDj3VnbmRq6Y3Z-GP8B_T9y9A |
| CitedBy_id | crossref_primary_10_7759_cureus_57037 crossref_primary_10_3390_jcm14030863 crossref_primary_10_1007_s10309_024_00709_1 crossref_primary_10_1111_epi_18034 crossref_primary_10_3238_arztebl_2025_Stellungnahme_KI_Medizin crossref_primary_10_3389_fmed_2025_1563003 crossref_primary_10_2196_63937 crossref_primary_10_1109_LSENS_2024_3515881 crossref_primary_10_1016_j_neuri_2024_100184 crossref_primary_10_1016_j_yebeh_2023_109486 crossref_primary_10_1111_epi_18138 |
| ContentType | Journal Article |
| Copyright | Copyright © 2023 British Epilepsy Association. Published by Elsevier Ltd. All rights reserved. |
| Copyright_xml | – notice: Copyright © 2023 British Epilepsy Association. Published by Elsevier Ltd. All rights reserved. |
| DBID | CGR CUY CVF ECM EIF NPM 7X8 |
| DOI | 10.1016/j.seizure.2023.06.002 |
| DatabaseName | Medline MEDLINE MEDLINE (Ovid) MEDLINE MEDLINE PubMed MEDLINE - Academic |
| DatabaseTitle | MEDLINE Medline Complete MEDLINE with Full Text PubMed MEDLINE (Ovid) MEDLINE - Academic |
| DatabaseTitleList | MEDLINE - Academic MEDLINE |
| Database_xml | – sequence: 1 dbid: NPM name: PubMed url: http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?db=PubMed sourceTypes: Index Database – sequence: 2 dbid: 7X8 name: MEDLINE - Academic url: https://search.proquest.com/medline sourceTypes: Aggregation Database |
| DeliveryMethod | no_fulltext_linktorsrc |
| Discipline | Medicine |
| EISSN | 1532-2688 |
| ExternalDocumentID | 37336056 |
| Genre | Journal Article |
| GroupedDBID | --- --K --M -RU .1- .FO .~1 0R~ 123 1B1 1P~ 1RT 1~. 1~5 4.4 457 4G. 53G 5VS 7-5 71M 8P~ 9JM 9JO AADFP AAEDT AAEDW AAGJA AAGUQ AAIKJ AAKOC AALRI AAOAW AAQFI AAQXK AATTM AAWTL AAXKI AAXLA AAXUO AAYWO ABBQC ABCQJ ABFNM ABIVO ABJNI ABMAC ABMZM ABOYX ABTEW ABWVN ABXDB ACDAQ ACGFS ACIEU ACRLP ACRPL ACVFH ACXNI ADBBV ADCNI ADEZE ADMUD ADNMO ADVLN AEBSH AEIPS AEKER AEUPX AEVXI AEXQZ AFCTW AFJKZ AFPUW AFRHN AFTJW AFXIZ AGCQF AGHFR AGQPQ AGRNS AGUBO AGWIK AGYEJ AHHHB AIEXJ AIGII AIIUN AIKHN AITUG AJRQY AJUYK AKBMS AKRLJ AKRWK AKYEP ALMA_UNASSIGNED_HOLDINGS AMRAJ ANKPU ANZVX ASPBG AVWKF AXJTR AZFZN BAWUL BKOJK BLXMC BNPGV CAG CGR COF CS3 CUY CVF DIK E3Z EBS ECM EFJIC EFKBS EIF EJD EO8 EO9 EP2 EP3 F5P FDB FEDTE FGOYB FIRID FNPLU FYGXN G-2 G-Q GBLVA HDW HMK HMO HVGLF HZ~ IHE IXB J1W KOM LX8 M29 M41 MO0 MOBAO N9A NPM O-L O9- OAUVE OK1 OKEIE OP~ OZT P-8 P-9 P2P PC. Q38 R2- RIG ROL RPZ SAE SCC SDF SDG SEL SES SEW SSB SSH SSN SSY SSZ T5K TR2 WUQ Z5R ~G- 7X8 ACLOT EFLBG ~HD |
| ID | FETCH-LOGICAL-c419t-c2e8f883b33708b379ffb54fa7c15d20cf0834bedc4924b97b59bcc35a0b223d2 |
| IEDL.DBID | 7X8 |
| ISICitedReferencesCount | 14 |
| ISICitedReferencesURI | http://www.webofscience.com/api/gateway?GWVersion=2&SrcApp=Summon&SrcAuth=ProQuest&DestLinkType=CitingArticles&DestApp=WOS_CPL&KeyUT=001055028600001&url=https%3A%2F%2Fcvtisr.summon.serialssolutions.com%2F%23%21%2Fsearch%3Fho%3Df%26include.ft.matches%3Dt%26l%3Dnull%26q%3D |
| ISSN | 1532-2688 |
| IngestDate | Sun Sep 28 00:40:20 EDT 2025 Mon Jul 21 05:56:57 EDT 2025 |
| IsDoiOpenAccess | false |
| IsOpenAccess | true |
| IsPeerReviewed | true |
| IsScholarly | true |
| Keywords | Deep learning Antiseizure medication Epilepsy Autonomous nervous system Explainable artificial intelligence Wearables Automatic seizure logging Tonic-clonic seizure |
| Language | English |
| License | Copyright © 2023 British Epilepsy Association. Published by Elsevier Ltd. All rights reserved. |
| LinkModel | DirectLink |
| MergedId | FETCHMERGED-LOGICAL-c419t-c2e8f883b33708b379ffb54fa7c15d20cf0834bedc4924b97b59bcc35a0b223d2 |
| Notes | ObjectType-Article-1 SourceType-Scholarly Journals-1 ObjectType-Feature-2 content type line 23 |
| OpenAccessLink | https://www.clinicalkey.com/#!/content/1-s2.0-S1059131123001553 |
| PMID | 37336056 |
| PQID | 2827924030 |
| PQPubID | 23479 |
| ParticipantIDs | proquest_miscellaneous_2827924030 pubmed_primary_37336056 |
| PublicationCentury | 2000 |
| PublicationDate | 2023-Aug 20230801 |
| PublicationDateYYYYMMDD | 2023-08-01 |
| PublicationDate_xml | – month: 08 year: 2023 text: 2023-Aug |
| PublicationDecade | 2020 |
| PublicationPlace | England |
| PublicationPlace_xml | – name: England |
| PublicationTitle | Seizure (London, England) |
| PublicationTitleAlternate | Seizure |
| PublicationYear | 2023 |
| SSID | ssj0009413 |
| Score | 2.439158 |
| Snippet | Objective seizure count estimates are crucial for ambulatory epilepsy management. Wearables have shown promise for the detection of tonic-clonic seizures but... |
| SourceID | proquest pubmed |
| SourceType | Aggregation Database Index Database |
| StartPage | 99 |
| SubjectTerms | Child Data Accuracy Electroencephalography - methods Epilepsy Humans Infant Retrospective Studies Seizures - diagnosis Seizures - drug therapy Wearable Electronic Devices |
| Title | Explainable AI for wearable seizure logging: Impact of data quality, patient age, and antiseizure medication on performance |
| URI | https://www.ncbi.nlm.nih.gov/pubmed/37336056 https://www.proquest.com/docview/2827924030 |
| Volume | 110 |
| WOSCitedRecordID | wos001055028600001&url=https%3A%2F%2Fcvtisr.summon.serialssolutions.com%2F%23%21%2Fsearch%3Fho%3Df%26include.ft.matches%3Dt%26l%3Dnull%26q%3D |
| hasFullText | |
| inHoldings | 1 |
| isFullTextHit | |
| isPrint | |
| link | http://cvtisr.summon.serialssolutions.com/2.0.0/link/0/eLvHCXMwpV1LS8NAEF7Uinjx_agvVvDY0CS76e56kSIWe2jpQaG3sk_wktTWKuqfd3azpSdBEJJADgvJzuzky87M9yF0YwvADYyapDBWJZQokShGnTdIWlhuM21oEJtgwyEfj8UobrjNY1nlMiaGQG0q7ffI2_BrwIQnj0vvpq-JV43y2dUoobGOGgSgjC_pYuMVW7igQR4ZFnWe5B3OVx08bd_R_fK1mHmmzJwECs-ws_ILygxfm97uf59zD-1EnIm7tWPsozVbHqCtQcykH6JvX3wXO6dwt48Bu-IP8PpwHx8aQ1z0Gka3uB96KXHlsK8oxXUn5mcLR1ZWDEGphWVp4PTizvXokLYPhsdwTFcdCkfouffwdP-YRCGGRNNMvCU6t9xxThQhLOWKMOGcKqiTTGeFyVPtAMhRZY2m8OpKMFUIpTUpZKoAfpj8GG2UVWlPEabMZsRJkjppaS4zmXNrhTLcFVp2nGqi6-W0TsDRffZClrZazCeriW2ik9o2k2nNyDEhntQRoNzZH0afo21v8rqI7wI1HCxze4k29TtM0OwqeBBch6PBD4nG0uM |
| linkProvider | ProQuest |
| openUrl | ctx_ver=Z39.88-2004&ctx_enc=info%3Aofi%2Fenc%3AUTF-8&rfr_id=info%3Asid%2Fsummon.serialssolutions.com&rft_val_fmt=info%3Aofi%2Ffmt%3Akev%3Amtx%3Ajournal&rft.genre=article&rft.atitle=Explainable+AI+for+wearable+seizure+logging%3A+Impact+of+data+quality%2C+patient+age%2C+and+antiseizure+medication+on+performance&rft.jtitle=Seizure+%28London%2C+England%29&rft.au=Halimeh%2C+Mustafa&rft.au=Jackson%2C+Michele&rft.au=Vieluf%2C+Solveig&rft.au=Loddenkemper%2C+Tobias&rft.date=2023-08-01&rft.eissn=1532-2688&rft.volume=110&rft.spage=99&rft_id=info:doi/10.1016%2Fj.seizure.2023.06.002&rft_id=info%3Apmid%2F37336056&rft_id=info%3Apmid%2F37336056&rft.externalDocID=37336056 |
| thumbnail_l | http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/lc.gif&issn=1532-2688&client=summon |
| thumbnail_m | http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/mc.gif&issn=1532-2688&client=summon |
| thumbnail_s | http://covers-cdn.summon.serialssolutions.com/index.aspx?isbn=/sc.gif&issn=1532-2688&client=summon |