Artifact classification¶
Hamid2020 (TUAR)Usage¶
neuralbench eeg artifact
Show config.yaml
# Copyright (c) Meta Platforms, Inc. and affiliates.
# All rights reserved.
#
# This source code is licensed under the license found in the
# LICENSE file in the root directory of this source tree.
data:
study:
source:
name: Hamid2020Tuar
split:
name: SklearnSplit
split_by: subject
valid_split_ratio: 0.2
test_split_ratio: 0.2
valid_random_state: 33
test_random_state: 33
target:
=replace=: true
name: LabelEncoder
event_types: Artifact
event_field: state
return_one_hot: true
aggregation: sum
allow_missing: true
trigger_event_type: Eeg
start: -1.0
duration: 3.0
stride: 1.0
stride_drop_incomplete: false
summary_columns: [state]
compute_class_weights: true
brain_model_output_size: &brain_model_output_size 5
trainer_config.monitor: val/f1_score_macro
trainer_config.mode: max
loss:
=replace=: true
name: BCEWithLogitsLoss
metrics: !!python/object/apply:neuralbench.defaults.metrics.get_classification_metric_configs
- *brain_model_output_size
- multilabel
Description¶
The goal of the artifact classification task is to predict whether 3-second EEG sliding windows contain one or mode of the following five artifact types [Hamid2020]:
eye movement (“eyem”)
chewing (“chew”)
shivers (“shiv”)
muscle artifact (“musc”)
electrode-related artifacts, i.e. electrode pop, electrostatic, or lead artifact (“elec”)
Of note, the seizure annotations available in this dataset are not used in this task.
Dataset Notes¶
Following the formulation of the clinical event classification task, we adopt a multilabel classification approach to better model overlapping events. See clinical event classification.
For this, we break down the combined artifact events (“eyem_musc”, “musc_elec”, “eyem_elec”, etc.) into two separate events with the same start and duration.
References¶
Hamid, Ahmed, et al. “The Temple University Artifact Corpus: An annotated corpus of EEG artifacts.” 2020 IEEE Signal Processing in Medicine and Biology Symposium (SPMB). IEEE, 2020.