Matches in UGent Biblio for { <https://biblio.ugent.be/publication/4246386#aggregation> ?p ?o. }
Showing items 1 to 28 of
28
with 100 items per page.
- aggregation classification "C3".
- aggregation creator person.
- aggregation creator person.
- aggregation creator person.
- aggregation date "2013".
- aggregation format "application/pdf".
- aggregation hasFormat 4246386.bibtex.
- aggregation hasFormat 4246386.csv.
- aggregation hasFormat 4246386.dc.
- aggregation hasFormat 4246386.didl.
- aggregation hasFormat 4246386.doc.
- aggregation hasFormat 4246386.json.
- aggregation hasFormat 4246386.mets.
- aggregation hasFormat 4246386.mods.
- aggregation hasFormat 4246386.rdf.
- aggregation hasFormat 4246386.ris.
- aggregation hasFormat 4246386.txt.
- aggregation hasFormat 4246386.xls.
- aggregation hasFormat 4246386.yaml.
- aggregation language "eng".
- aggregation rights "I have transferred the copyright for this publication to the publisher".
- aggregation subject "Technology and Engineering".
- aggregation title "Building a patient-specific seizure detector without expert input using user triggered active learning strategies".
- aggregation abstract "Purpose: Patient-specific seizure detectors outperform general seizure detectors, but building them requires lots of consistently marked electroencephalogram (EEG) of a single patient, which is expensive to gather. This work presents a method to bring general seizure detectors up to par with patient-specific seizure detectors without expert input. The user/patient is only required to push a button in case of a false alarm and/or missed seizure. Method: For the experiments the 'CHB-MIT Scalp EEG Database' was used, which contains pre-surgically recorded EEG of 24 patients. The seizure detector used is based on (Buteneers et al. Epilepsy Research 2012:(in press)) combined with the preprocessing technique presented in (Shoeb et al. Epilepsy & Behavior 2004;5:483-598). Button presses mark the corresponding data and add it to the training set of the system. The performance is evaluated using leave-one-hour-out cross-validation to attain statistically relevant results. Results: For the patient-specific seizure detector 34(32)% (average(standard deviation)) of the detections are false, 8(14)% of the seizures are missed and a detection delay of 11(10)s is reached. The general seizure detector achieves: 86(89)%, 28(41)% and -35(82)s, respectively. Adding only false positives, the patient specific performance is achieved in 9 of the 24 patients. Adding missed seizures allows the patient-specific performance to be reached in 21 patients (about 90%). Conclusion: This work shows that in order to build a patient-specific seizure detector, no patient-specific EEG data is required for up to 90% of the patients using the presented technique.".
- aggregation authorList BK275773.
- aggregation aggregates 4246430.
- aggregation isDescribedBy 4246386.
- aggregation similarTo LU-4246386.