MAIT v1.0.0: First Official Release of the Medical Artificial Intelligence Toolbox

We are happy to announce the first official release of MAIT (Medical Artificial Intelligence Toolbox) v1.0.0. This open-source Python framework is designed to streamline machine learning workflows for tabular datasets, with a primary focus on binary classification, survival modeling, and regression analyses.

Key Features

Unified Machine Learning Framework

• Supports binary classification, regression, and survival models in a single, cohesive pipeline.
• Built with TRIPOD+AI compliance to ensure transparency and reproducibility.

Advanced Model Explainability

• Integrated tools like SHAP (SHapley Additive exPlanations) for feature importance analysis and model interpretability.
• Enhanced visualizations for mixed data types, including SHAP plots for categorical features.

Comprehensive Data Handling

• Preprocessing solutions for high-dimensional datasets, missing values, and class imbalances.
• Automatic data quality checks and robust imputation methods.

Survival-to-Binary Translation

• Novel functionality to translate cumulative hazard curves into binary classifications, enabling intuitive comparisons across model types.

External Validation and Benchmarking

• Seamless support for external dataset validation and multi-metric performance evaluations.

Workflow Flexibility

• Customizable configurations for both beginners and advanced users.
• GPU support for faster processing.

Documentation and Tutorials

To help users get started quickly, MAIT includes a detailed documentation portal with step-by-step tutorials covering practical use cases such as:

1. Prediction of antimicrobial resistance for Azithromycin and Ciprofloxacin.
2. Prediction of dementia outcomes.
3. Diagnosis of breast cancer.

Get Started

• GitHub Repository: PERSIMUNE/MAIT
• Documentation: MAIT Docs

We’re excited to share MAIT with the research community and invite feedback to help us improve and expand its capabilities.