Dataset

This dataset was created by M R Khan Dipu 393

Contents

Epilepsy Detection Dataset for Federated Deep Learning Overview This dataset contains comprehensive EEG-derived features collected for the purpose of developing federated deep learning models aimed at epilepsy detection. The dataset comprises 289,010 records, each representing an EEG recording segment annotated with various time-domain, frequency-domain, wavelet transform, nonlinear, seizure-specific, and demographic features.

The primary objective of this dataset is to facilitate research in real-time epilepsy detection while ensuring data privacy through federated learning techniques. The dataset includes both multi-class labels (Normal, Pre-Seizure, Seizure, Post-Seizure) and seizure type classifications (Normal, Generalized Seizure, Focal Seizure).

Data Collection Details The data was collected in real-time from multiple clinical EEG recording systems across diverse hospitals. Signals were recorded at standard sampling rates with consistent preprocessing protocols to ensure data uniformity and high-quality feature extraction. Subjects include patients aged between 1 and 90 years, ensuring a broad demographic representation.

Features Description The dataset includes 75 features categorized into six main groups:

1. Time-Domain Features (15 Features) These features capture statistical properties of the EEG signal in the time domain, providing valuable insights into the signal’s amplitude variations and temporal patterns.

Mean_EEG_Amplitude: Average amplitude across EEG segments. EEG_Std_Dev: Standard deviation of the EEG signal, reflecting variability. EEG_Skewness: Measures asymmetry of the EEG signal distribution. EEG_Kurtosis: Degree of peakedness or flatness in the EEG signal. Zero_Crossing_Rate: Frequency of signal sign changes. Root_Mean_Square: Signal energy magnitude indicator. Peak_to_Peak_Amplitude: Difference between maximum and minimum amplitude. Signal_Energy: Energy content of EEG segments. Variance_of_EEG_Signals: Variability of signal amplitude. Interquartile_Range: Range between the 25th and 75th percentile amplitudes. Auto_Correlation_of_EEG_Signals: Similarity between signal values at different lags. Cross_Correlation_Between_Channels: Measures inter-channel dependencies. Hjorth_Mobility: Frequency-dependent signal descriptor. Hjorth_Complexity: Complexity of EEG waveform changes. Line_Length_Feature: Cumulative length of the EEG waveform trajectory. 2. Frequency-Domain Features (10 Features) Frequency features highlight spectral content and distribution, essential for capturing seizure-related oscillations.

Delta_Band_Power: Power within the delta frequency range. Theta_Band_Power: Theta band power variations. Alpha_Band_Power: EEG activity in the alpha band. Beta_Band_Power: Beta frequency energy (notable for cognitive activity). Gamma_Band_Power: High-frequency brain activity measures. Low_to_High_Frequency_Power_Ratio: Indicator of frequency band shifts during seizures. Power_Spectral_Density: Power distribution across frequencies. Spectral_Edge_Frequency: Frequency below which a certain percentage of power is contained. Spectral_Entropy: Signal complexity in the frequency domain. Fourier_Transform_Features: Global frequency representation through Fourier analysis. 3. Wavelet Transform Features (5 Features) Wavelet-based features capture transient events and non-stationary patterns in the EEG signal.

Wavelet_Entropy: Information content using wavelet decomposition. Wavelet_Energy: Energy derived from wavelet coefficients. Discrete_Wavelet_Transform: Detailed frequency analysis at different scales. Continuous_Wavelet_Transform: Continuous frequency-time representation. Wavelet_Based_Shannon_Entropy: Entropy-based wavelet feature. 4. Nonlinear Features (10 Features) Nonlinear measures provide insights into the dynamic and chaotic nature of EEG signals.

Sample_Entropy: Complexity and unpredictability of the signal. Approximate_Entropy: Regularity of signal fluctuations. Shannon_Entropy: Signal randomness indicator. Permutation_Entropy: Complexity through sequence ordering. Lyapunov_Exponent: Sensitivity to initial conditions (chaotic behavior). Hurst_Exponent: Long-term memory effect measurement. Detrended_Fluctuation_Analysis: Scale-dependent correlations. Higuchi_Fractal_Dimension: Signal complexity measure using fractal geometry. Katz_Fractal_Dimension: Alternative fractal dimension metric. Lempel_Ziv_Complexity: Signal compressibility and complexity measure. 5. Seizure-Specific Features (6 Features) Features tailored to capture seizure onset, duration, and recovery patterns.

Seizure_Duration: Duration of seizure episodes (in seconds). Pre_Seizure_Pattern: Indicators preceding seizure onset. Post_Seizure_Recovery: Recovery patterns after seizure termination. Seizure_Frequency_Per_Hour: Number of seizures occurring per hour. Interictal_Spike_Rate: Frequency of spikes between se...

Dataset

This dataset was created by Rajorshi Chaudhuri

Released under CC BY-SA 4.0

Contents

Source: [9].

Dataset

This dataset was created by Abhi

Released under CC0: Public Domain

Contents

Univ. of Bonn’ and ‘CHB-MIT Scalp EEG Database’ are publically available datasets which are the most sought after amongst researchers. Bonn dataset is very small compared to CHB-MIT. But still researchers prefer Bonn as it is in simple '.txt' format. The dataset being published here is a preprocessed form of CHB-MIT. The dataset is available in '.csv' format.

Dataset

This dataset was created by Johar M. Ashfaque

Contents

Dataset

This dataset was created by Johar M. Ashfaque

Contents

Grand Mal Seizure - Patterns & Predictions

Context

This data set is 12 months of seizure frequency data of a 20 year old male that lives in Iowa City, Iowa. He sufferers from grand mal seizures that result from Tubular Sclerosis that affects his brain. This is my nephew Michael, and I love him, and I want to access the smartest and best data resources to help provide insight and maybe support that can give him some relief.

I am providing this data in hopes of creating a contest that can help identify and predict a pattern to his seizures based on this frequency data and state (sleep/awake), and environmental data including weather, allergens, temperature, and other potential sources.

About Tubular Sclerosis

Tubular Sclerosis is an uncommon genetic disorder that causes noncancerous (benign) tumors — unexpected overgrowths of normal tissue — to develop in many parts of the body.

Context

This data was extracted from the CATS database dump released by the Justice Department under a FOIA request as part of the Pulitzer Center's Taken project.

Content

The data is available as CSV and JSON.

Acknowledgements

Data extraction was by performed by Maptian and supported by the Pulitzer Center.

Inspiration

We're hoping to encourage analysis of this data by reporters and experts in the field. If you are a newsroom interested in joining the Taken project, please get in touch.

Seizure method codes

A: Adoption F: Warrantless/PC C: Search Warrent H: Judgement I: Incident Arrest D: Seizure Warrent N: Consent O: Other B: Indictment J: Civil Complaint Q: Warrent - Federal Seizure P: Warrent - State Seizure R: Warrent - Federal Search and Seizure

Asset type codes

VH: Vehicle CA: Cash AL: Alcohol CH: Chemicals FI: Financial Instrument WE: Firearms RP: Real Property EL: Electronic Equipment AM: Ammunition OT: Other

Judicial district codes

ALM: Alabama, Middle ALN: Alabama, Northern ALS: Alabama, Southern AK: Alaska AZ: Arizona ARE: Arkansas, Eastern ARW: Arkansas, Western CAC: California, Central CAE: California, Eastern CAN: California, Northern CAS: California, Southern CO: Colorado CT: Connecticut DE: Delaware FLM: Florida, Middle FLN: Florida, Northern FLS: Florida, Southern GAM: Georgia, Middle GAN: Georgia, Northern GAS: Georgia, Southern HI: Hawaii ID: Idaho ILC: Illinois, Central ILN: Illinois, Northern ILS: Illinois, Southern INN: Indiana, Northern INS: Indiana, Southern IAN: Iowa, Northern IAS: Iowa, Southern KS: Kansas KYE: Kentucky, Eastern KYW: Kentucky, Western LAE: Louisiana, Eastern LAM: Louisiana, Middle LAW: Louisiana, Western ME: Maine MD: Maryland MA: Massachusetts MIE: Michigan, Eastern MIW: Michigan, Western MN: Minnesota MSN: Mississippi, Northern MSS: Mississippi, Southern MOE: Missouri, Eastern MOW: Missouri, Western MT: Montana NE: Nebraska NV: Nevada NH: New Hampshire NJ: New Jersey NM: New Mexico NYE: New York, Eastern NYN: New York, Northern NYS: New York, Southern NYW: New York, Western NCE: North Carolina, Eastern NCM: North Carolina, Middle NCW: North Carolina, Western ND: North Dakota OHN: Ohio, Northern OHS: Ohio, Southern OHE: Oklahoma, Eastern OHN: Oklahoma, Northern OHW: Oklahoma, Western OR: Oregon PEE: Pennsylvania, Eastern PEM: Pennsylvania, Middle PEW: Pennsylvania, Western RI: Rhode Island SC: South Carolina SD: South Dakota TNE: Tennessee, Eastern TNM: Tennessee, Middle TNW: Tennessee, Western TXE: Texas, Eastern TXN: Texas, Northern TXS: Texas, Southern TXW: Texas, Western UT: Utah VT: Vermont VAE: Virginia, Eastern VAW: Virginia, Western DC: Washington, D.C. WAW: Washington, Eastern WAW: Washington, Western WVN: West Virginia, Northern WVS: West Virginia, Southern WIE: Wisconsin, Eastern WIW: Wisconsin, Western WY: Wyoming

Context

This Dataset is a time-varying criminal network that is repeatedly disturbed by police forces. The CAVIAR investigation lasted two years and ran from 1994 to 1996. The operation brought together investigation units of the Montréal police and the Royal Canadian Mounted Police of Canada. During this two-year period, 11 wiretap warrants, valid for a period of about two months each, were obtained (the 11 matrices contained in phase1.csv, phase2.csv, correspond to these eleven, two-month wiretap phases).

Content

This case is interesting because, unlike other investigative strategies, the mandate of the CAVIAR project was to seize the drugs without arresting the perpetrators. During this period, imports of the trafficking network were hit by the police on eleven occasions. The arrests took place only at the end of the investigation. Monetary losses for traffickers were estimated at 32 million dollars. Eleven seizures took place throughout the investigation. Some phases included no seizures, and others included multiple.

The following summarizes the seizures:

Phase 4 1 seizure $2,500,000 300 kg of marijuana Phase 6 3 seizures $1,300,000 2 x 15 kg of marijuana + 1 x 2 kg of cocaine Phase 7 1 seizure $3,500,000 401 kg of marijuana Phase 8 1 seizure $360,000 9 kg of cocaine Phase 9 2 seizures $4,300,000 2 kg of cocaine + 1 x 500 kg marijuana Phase 10 1 seizure $18,700,000 2200 kg of marijuana Phase 11 2 seizures $1,300,000 12 kg of cocaine + 11 kg of cocaine

This case offers a rare opportunity to study a criminal network in upheaval from police forces. This allows us to analyze changes in the network structure and to survey the reaction and adaptation of the participants while they were subjected to an increasing number of distressing constraints.

The network consists of 110 (numbered) players. Players 1-82 are the traffickers. Players 83-110 are the non-traffickers (financial investors; accountants; owners of various importation businesses, etc.). Initially, the investigation targeted Daniel Serero, the alleged mastermind of a drug network in downtown Montréal, who attempted to import marijuana to Canada from Morocco, transiting through Spain. After the first seizure, happening in Phase 4, traffickers reoriented to cocaine import from Colombia, transiting through the United States.

According to the police, the role of 23 of the players in the “Serero organization" are the following, listed by name (unique id):

Daniel Serero (n1) : Mastermind of the network.

Pierre Perlini (n3) : Principal lieutenant of Serero, he executes Serero's instructions.

Alain (n83) and Gérard (n86) Levy : Investors and transporters of money.

Wallace Lee (n85) : Takes care of financial affairs (accountant).

Gaspard Lino (n6): Broker in Spain.

Samir Rabbat (n11): Provider in Morocco.

Lee Gilbert (n88): A trusted man of Wallace Lee (became an informer after the arrest).

Beverly Ashton (n106): Spouse of Lino, transports money and documents.

Antonio Iannacci (n89): Investor.

Mohammed Echouafni (n84): Moroccan investor.

Richard Gleeson (n5), Bruno de Quinzio (n8) and Gabrielle Casale (n76) : Charged with recuperating the marijuana.

Roderik Janouska (n77): Individual with airport contacts.

Patrick Lee (n87): Investor.

Salvatore Panetta (n82): Transport arrangements manager.

Steve Cunha (n96): Transport manager, owner of a legitimate import company (became an informer after the arrest).

Ernesto Morales (n12): Principal organizer of the cocaine import, an intermediary between the Colombians and the Serero organization.

Oscar Nieri (n17): The handyman of Morales.

Richard Brebner (n80): Was transporting the cocaine from the US to Montréal.

Ricardo Negrinotti (n33): Was taking possession of the cocaine in the US to hand it to Brebner.

Johnny Pacheco (n16): Cocaine provider.

In the data files (phase1.csv, phase2.csv, ), you will find matrices that report the number of wiretapped correspondences between the above players in the network, where players are identified by their unique id. You will be analyzing this time-varying network, giving a rough sketch of its shape, its evolution and the role of the actors in it.

You can also load this dataset directly from the internet (which is slightly more convenient when using Colab) using the following Python code:

import pandas as pd import networkx as nx phases = {} G = {} for i in range(1,12): var_name = "phase" + str(i) file_name = "https://raw.githubusercontent.com/ragini30/Networks-Homework/main/" + var_name + ".csv" phases[i] = pd.read_csv(file_name, index_col = ["players"]...