Home

Pixels Wiki

Pixels is a Databricks Industry Solutions accelerator for medical imaging (DICOM). It indexes imaging files stored in cloud storage or Unity Catalog Volumes into a Delta table (the object catalog), extracts DICOM headers as queryable VARIANT metadata, and on top of that ships DICOMweb apps, an OHIF viewer, Vista3D model serving, a Lakeview dashboard and an AI/BI Genie space.

This wiki is the reference for the Python library (dbx.pixels). For deploying the full stack, see docs/INSTALL.md.

The core flow

Every Pixels pipeline follows the same three steps:

from dbx.pixels import Catalog
from dbx.pixels.dicom import DicomMetaExtractor

catalog = Catalog(spark, table="main.pixels_solacc.object_catalog",
                  volume="main.pixels_solacc.pixels_volume")          # 1. configure

catalog_df = catalog.catalog("/Volumes/main/pixels_solacc/landing/")   # 2. discover files
meta_df = DicomMetaExtractor(catalog, permissive=True,
                             remove_un_tags=True).transform(catalog_df)  # 3. extract DICOM headers

catalog.save(meta_df)                                                 # persist to Delta

Step	Class	What it does
Discover	Catalog.catalog()	Lists files (batch, streaming or managed file events), optionally unzips archives, derives path columns
Extract	DicomMetaExtractor	Reads each DICOM header in parallel and adds meta, study_uid and series_uid columns
Persist	Catalog.save()	Writes the result to the object catalog Delta table, liquid-clustered by study_uid

Pages

• Ingestion: the Catalog class, how batch, streaming and managed file events differ, when to use each, and every parameter
• Metadata Extraction: the DicomMetaExtractor class, its parameters, output format, and how to query the meta column
• Release Notes: what changed in each release, including unreleased changes on main

More pages are planned for anonymization, redaction, thumbnails, DICOMweb, OHIF/MONAI and NIfTI overlays.