What clinical prediction tasks are predefined in pyhealth?

It includes 20+ predefined tasks such as mortality prediction, length of stay, 30-day readmission, drug recommendation, and sleep staging.

Can I use pyhealth for medical code mapping?

Yes, it includes specialized CrossMap and InnerMap tools to translate between various coding systems such as ICD-9, ICD-10, NDC, RxNorm, and ATC.

What datasets does the pyhealth skill support?

It supports over 10 major healthcare datasets, including MIMIC-III, MIMIC-IV, eICU, OMOP, and various physiological signal datasets like SleepEDF.

How does pyhealth handle data leakage in clinical settings?

It follows best practices by providing built-in functions to split data by patient, ensuring that data from the same individual never appears in both training and test sets.

Does this skill help with model interpretability?

Absolutely. It supports healthcare-specific models like RETAIN and AdaCare that are specifically designed for clinical interpretability and attention visualization.

PyHealth - Healthcare AI Toolkit

Name: PyHealth - Healthcare AI Toolkit
Author: Microck

byMicrock

•

Ciencia de Datos y ML

Develops and deploys specialized machine learning models for healthcare using clinical datasets and medical coding systems.

Acerca de

PyHealth is a comprehensive toolkit designed to streamline the complex workflow of healthcare AI, offering a modular 5-stage pipeline for data loading, task definition, model selection, training, and deployment. It provides out-of-the-box support for major clinical datasets like MIMIC-IV and eICU, automates translation between medical coding systems like ICD and RxNorm, and includes specialized deep learning architectures such as RETAIN and SafeDrug. This skill is ideal for researchers and developers working on electronic health records (EHR), physiological signal analysis, and clinical prediction tasks where data standardization and model interpretability are paramount.

Características Principales

Automated medical code translation between systems like ICD-9/10, NDC, and ATC
Modular 5-stage pipeline for clinical machine learning workflows
Advanced clinical evaluation tools for fairness, calibration, and uncertainty quantification
81 GitHub stars
Implementation of 30+ specialized models for healthcare prediction and interpretability
Standardized interfaces for 10+ healthcare datasets including MIMIC-IV and OMOP

Casos de Uso

Classifying physiological signals such as EEG for sleep staging and disorder diagnosis
Recommending safe medications while accounting for drug-drug interactions
Predicting patient mortality or hospital readmission risks using EHR data

Acerca de

Características Principales

Automated medical code translation between systems like ICD-9/10, NDC, and ATC
Modular 5-stage pipeline for clinical machine learning workflows
Advanced clinical evaluation tools for fairness, calibration, and uncertainty quantification
81 GitHub stars
Implementation of 30+ specialized models for healthcare prediction and interpretability
Standardized interfaces for 10+ healthcare datasets including MIMIC-IV and OMOP

Casos de Uso

Classifying physiological signals such as EEG for sleep staging and disorder diagnosis
Recommending safe medications while accounting for drug-drug interactions
Predicting patient mortality or hospital readmission risks using EHR data