Privacy-Preserving Early Detection of Epileptic Seizures in Videos

Mehta, Deval; Sivathamboo, Shobi; Simpson, Hugh; Kwan, Patrick; O’Brien, Terence; Ge, Zongyuan

doi:10.1007/978-3-031-43904-9_21

Cited by 3 publications

(6 citation statements)

References 28 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…50 They have found use as robust classifiers within the field of seizure analysis, though they lack a built-in method of dealing with unknown values or temporal data. 23,32,51,52,53,55,56,57 Tree-based methods Figure 3C Tree-based alogrithms are simple iterative methods often used in seizure analysis. 14,76 At each node of the tree, a feature is used to split the dataset into two groups based on whether they are above or below a threshold.…”

Section: Algorithm Descriptionmentioning

confidence: 99%

“…46 Because images can be expressed as a sequence of pixels or image patches, transformers can be applied to CV both in conjunction with CNNs 47 and independent of CNNs. 48,57 In the context of seizure semiology, transformers can potentially provide more accurate analyses of seizures in comparison to the CNN-based models employed by most recent papers. However, transformers generally require much more training data.…”

Section: Algorithm Descriptionmentioning

confidence: 99%

“…Gelastic, dacrystic or without an ictal cry, and a convulsive clonic phase. 55 Several studies have used OF-based methods to detect the periodic movements associated with clonic activity, 23,24,25,28,32,56,57,58 resulting in a sensitivity of >95% for clonic and GTC seizures in adults and children. 24,25,32,58 OF has been combined with other modalities to enhance seizure detection; combination with electroencephalography (EEG) outperforms either EEG or video detection alone in detecting nocturnal TC seizures in children.…”

Section: Fer Logistic Regressionmentioning

confidence: 99%

“…58 Thermal imaging also has the potential to be unidentifiable, allowing sharing of datasets without privacy concerns. 13,57 A widely generalizable seizure detection system must also be able to detect seizures that occur at home and in public places. Despite this, the vast majority of models were trained only on EMU datasets, and only a limited number of studies have used non-EMU videos.…”

Section: Improving Datasetsmentioning

confidence: 99%

“…also created an end‐to‐end model to analyze multiple frames, significantly improving the performance. Nocturnal BTC seizures have been detected in the infrared (IR) spectrum using thermophile sensor arrays and passive IR sensors to detect 1.6–3‐Hz periodic movements 13,51,57,60 . These devices, unlike IR illuminated cameras, produce unidentifiable images, protecting patient privacy while maintaining performance; models achieved up to 93.1% accuracy on patient data and accuracy of 99% on actor‐simulated seizures.…”

Section: Clinical Application In Epilepsymentioning

confidence: 99%

See 4 more Smart Citations

Computer vision for automated seizure detection and classification: A systematic review

Brown,

Boyne,

Hassan

et al. 2024

Epilepsia

View full text Add to dashboard Cite

Computer vision (CV) shows increasing promise as an efficient, low‐cost tool for video seizure detection and classification. Here, we provide an overview of the fundamental concepts needed to understand CV and summarize the structure and performance of various model architectures used in video seizure analysis. We conduct a systematic literature review of the PubMed, Embase, and Web of Science databases from January 1, 2000 to September 15, 2023, to identify the strengths and limitations of CV seizure analysis methods and discuss the utility of these models when applied to different clinical seizure phenotypes. Reviews, nonhuman studies, and those with insufficient or poor quality data are excluded from the review. Of the 1942 records identified, 45 meet inclusion criteria and are analyzed. We conclude that the field has shown tremendous growth over the past 2 decades, leading to several model architectures with impressive accuracy and efficiency. The rapid and scalable detection offered by CV models holds the potential to reduce sudden unexpected death in epilepsy and help alleviate resource limitations in epilepsy monitoring units. However, a lack of standardized, thorough validation measures and concerns about patient privacy remain important obstacles for widespread acceptance and adoption. Investigation into the performance of models across varied datasets from clinical and nonclinical environments is an essential area for further research.

show abstract