seizurekit 0.1.0

A Python toolkit for seizure detection evaluation and analysis.

SeizureKit

A Python toolkit for the robust evaluation and analysis of seizure detection models.

Overview

SeizureKit provides a suite of tools designed for researchers and developers working on seizure detection algorithms. It simplifies the process of evaluating model performance at both the epoch and event levels, offering standardized metrics, statistical tests, and post-processing utilities.

Key Features

• Epoch-Level Metrics: Calculate standard classification metrics like F1-score, precision, recall, and specificity on a per-window (epoch) basis.
• Event-Level Metrics: Evaluate performance based on seizure events, measuring detection latency, F1-score, and other event-based statistics.
• Post-Processing: Includes temporal smoothing and hysteresis thresholding to improve the stability and accuracy of predictions.
• Statistical Analysis: Perform statistical significance testing with tools like Bootstrapping and DeLong's test for comparing ROC curves.
• Visualization (Upcoming): Tools for plotting ground truth, model probabilities, and predictions.

Installation

You can install SeizureKit directly from PyPI (once published):

pip install seizurekit

Alternatively, for the latest development version, you can install it from the source:

git clone https://github.com/Mr-TalhaIlyas/seizurekit.git
cd seizurekit
pip install -e .

Usage

SeizureKit can be used both as a command-line tool for quick evaluations and as a library in your Python projects for more complex workflows.

As a Command-Line Tool

After installation, you can use the seizurekit command to evaluate your model's predictions. You need a ground truth file and a model probabilities file (either .npy or .csv).

# Display help and all available options
seizurekit --help

# Example: Run a full evaluation
seizurekit path/to/ground_truth.npy path/to/model_probabilities.npy --smoothing_window 5 --hyst_high 0.7 --hyst_low 0.3

As a Python Library

Integrate SeizureKit's functions directly into your analysis scripts.

import numpy as np
from seizurekit.eval_epochs import calculate_classification_metrics
from seizurekit.eval_events import calculate_event_level_metrics

# 1. Load or create your data
y_true = np.array([0, 0, 1, 1, 1, 0, 0, 0, 1, 1, 0, 0, 0, 1, 1, 1, 1])
y_prob_1 = np.array([0.1, 0.2, 0.8, 0.9, 0.4, 0.3, 0.1, 0.4, 0.85, 0.95, 0.2, 0.1, 0.3, 0.99, 0.98, 0.97, 0.5])
y_prob_2 = np.array([0.05, 0.15, 0.75, 0.85, 0.35, 0.25, 0.05, 0.35, 0.8, 0.9, 0.15, 0.05, 0.25, 0.95, 0.9, 0.92, 0.45])

all_targets = y_true
all_probs = {'my_model': y_prob_1,
              'my_model_2': y_prob_2}
eval_config = {'window_duration': 10, 'detection_tolerance': 30}

# 2. Run epoch-level evaluation
print("--- Epoch-Level Metrics ---")
epoch_results = calculate_classification_metrics(
    all_probs,
    all_targets,
    eval_config,
    smoothing_window=3,
    hysteresis_high=0.5,
    hysteresis_low=0.5
)

# 3. Run event-level evaluation
print("\n--- Event-Level Metrics ---")
event_results = calculate_event_level_metrics(
    all_probs,
    all_targets,
    eval_config,
    smoothing_window=3,
    hysteresis_high=0.5,
    hysteresis_low=0.5
)

Running above code produces output below;

--- Epoch-Level Metrics ---
========================================================================================================================
Modality             Recall (CI)          Precision (CI)       F1-Score (CI)        Specificity NPV        AUROC (CI)           AP (CI)             
------------------------------------------------------------------------------------------------------------------------
my_model             0.941 [0.824,1.000]     0.948 [0.872,1.000]     0.941 [0.820,1.000]   1.000      0.889      1.000 [1.000,1.000]     1.000 [1.000,1.000]
my_model_2           0.882 [0.706,1.000]     0.906 [0.837,1.000]     0.882 [0.706,1.000]   1.000      0.800      1.000 [1.000,1.000]     1.000 [1.000,1.000]
========================================================================================================================

Modality             TN         FP         FN         TP        
my_model             8          0          1          8         
my_model_2           8          0          2          7         
================================================================================

Total Duration: 0.05 hours
Modality             False Alarms/h 
my_model             0.00            FPR/h
my_model_2           0.00            FPR/h
================================================================================

--- Event-Level Metrics ---
========================================================================================================================
Event-Level Detection Metrics:
------------------------------------------------------------------------------------------------------------------------
Modality             | #Events  | TP    | FN    | FP    | Sens   | Lat (s)  | FA/h   | Cov.   | Burden (min/h) 
------------------------------------------------------------------------------------------------------------------------
my_model             | 3        | 3     | 0     | 0     | 1.0000 | 0.0000   | 0.0000 | 0.8889 | 28.2353        
my_model_2           | 3        | 3     | 0     | 0     | 1.0000 | 0.0000   | 0.0000 | 0.7778 | 24.7059        
========================================================================================================================
Coverage(Cov.): Fraction of true seizure time detected by AI.
Burden: Time AI believes seizure is occurring, scaled to minutes/hour.
========================================================================================================================

Contributing

Contributions are welcome! If you have suggestions for improvements or new features, feel free to open an issue or submit a pull request.

1. Fork the repository.
2. Create a new branch (git checkout -b feature/your-feature-name).
3. Make your changes.
4. Commit your changes (git commit -m 'Add some feature').
5. Push to the branch (git push origin feature/your-feature-name).
6. Open a pull request.

License

This project is licensed under the MIT License. You should create a LICENSE file in your project root with the MIT