Does PyHealth support model interpretability for clinical use?

Absolutely. It includes tools for attention visualization (RETAIN), calibration (Platt scaling), and uncertainty quantification to ensure models are safe for clinical deployment.

Can this skill help with medical coding translation?

Yes, it includes specialized tools for mapping between diverse medical coding systems such as ICD-9/10, NDC, RxNorm, ATC, and CCS.

What healthcare datasets does the PyHealth skill support?

PyHealth provides standardized interfaces for over 10 major datasets, including MIMIC-III, MIMIC-IV, eICU, OMOP, and various EEG/ECG signal datasets.

Are there healthcare-specific models included in this skill?

Yes, the skill supports 33+ models, including general architectures like Transformers and GNNs, as well as healthcare-specific models like RETAIN, SafeDrug, and StageNet.

Which clinical prediction tasks are predefined in PyHealth?

It features 20+ predefined tasks including mortality prediction, hospital readmission, length of stay, drug recommendation, and sleep staging.

PyHealth AI Toolkit

Name: PyHealth AI Toolkit
Author: x-cmd

byx-cmd

•

데이터 과학 및 ML

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

소개

PyHealth is a comprehensive library designed to streamline healthcare AI development, providing a modular 5-stage pipeline for building, testing, and deploying clinical prediction models. This skill enables developers to work efficiently with electronic health records (EHR), physiological signals, and medical images by offering standardized access to major datasets like MIMIC-IV and eICU. It simplifies complex tasks such as medical coding translation (ICD, NDC, ATC) and implements specialized healthcare architectures like RETAIN and SafeDrug to ensure models are interpretable and clinically relevant.

주요 기능

Automated medical coding translation across ICD, NDC, RxNorm, and ATC
Advanced evaluation metrics for clinical fairness, calibration, and interpretability
33+ healthcare-specific models for clinical prediction and recommendation
20+ predefined tasks for mortality, readmission, and length of stay
Standardized loading for 10+ clinical datasets including MIMIC and eICU
8 GitHub stars

사용 사례

Predicting patient mortality and 30-day hospital readmission risks
Classifying sleep stages and detecting seizures from EEG physiological signals
Developing medication recommendation systems with drug-drug interaction constraints

소개

주요 기능

Automated medical coding translation across ICD, NDC, RxNorm, and ATC
Advanced evaluation metrics for clinical fairness, calibration, and interpretability
33+ healthcare-specific models for clinical prediction and recommendation
20+ predefined tasks for mortality, readmission, and length of stay
Standardized loading for 10+ clinical datasets including MIMIC and eICU
8 GitHub stars

사용 사례

Predicting patient mortality and 30-day hospital readmission risks
Classifying sleep stages and detecting seizures from EEG physiological signals
Developing medication recommendation systems with drug-drug interaction constraints