pyhealth.tasks.bmd_hs_disease_classification

class pyhealth.tasks.bmd_hs_disease_classification.BMDHSDiseaseClassification

Bases: BaseTask

Multi-label classification task for heart valve diseases.

This task classifies heart sound recordings into multiple disease categories: - AS (Aortic Stenosis) - AR (Aortic Regurgitation) - MR (Mitral Regurgitation) - MS (Mitral Stenosis)

Each patient can have multiple diseases simultaneously (multi-label).

The task also provides access to patient metadata including: - Age - Gender - Smoker status (binary) - Living environment (urban/rural, binary)

And 8 heart sound recording filenames.

task_name

The name of the task, set to “BMDHSDiseaseClassification”.

Type:

str

input_schema

The input schema specifying the required input format. Contains: - “recording_1” through “recording_8”: “audio” - “age”: “float” - “gender”: “categorical” - “smoker”: “binary” - “lives”: “binary”

Type:

Dict[str, str]

output_schema

The output schema specifying the output format. Contains: - “diagnosis”: “multilabel”

Type:

Dict[str, str]

Examples

>>> from pyhealth.datasets import BMDHSDataset
>>> from pyhealth.tasks import BMDHSDiseaseClassification
>>> dataset = BMDHSDataset(root="/path/to/bmd_hs")
>>> task = BMDHSDiseaseClassification()
>>> samples = dataset.set_task(task)

task_name: str = 'BMDHSDiseaseClassification'

input_schema: Dict[str, str] = {'age': 'regression', 'gender': 'binary', 'lives': 'binary', 'recording_1': 'audio', 'recording_2': 'audio', 'recording_3': 'audio', 'recording_4': 'audio', 'recording_5': 'audio', 'recording_6': 'audio', 'recording_7': 'audio', 'recording_8': 'audio', 'smoker': 'binary'}

output_schema: Dict[str, str] = {'diagnosis': 'multilabel'}

pre_filter(df)

Return type:

LazyFrame