extracting-lab-tables

To extract lab tables from scanned images, apply on-device OCR using Tesseract or PaddleOCR backends to capture text and positioned words. Geometry-based alignment then reconstructs the detected words into structured test, value, unit, reference range, and abnormal flag columns.

Yes, you can process delimited CSV and TSV lab exports by parsing and classifying columns into clinical result fields. The skill identifies and redacts patient-identifying fields to produce PHI-safe structured rows for downstream analysis.

Yes, table reconstruction preserves reference ranges, units, and abnormal flags. Detected words from OCR are organized into structured columns specifically designed to maintain the information needed for clinical analysis and LOINC or UCUM mapping.

The best way to prepare scanned CBC and CMP reports for FHIR workflows is applying OCR to extract text, reconstructing tables into clinical rows, and mapping values to LOINC and UCUM standards. This produces structured rows ready for FHIR Observation creation.

Yes, on-device OCR with positioned word output is required for lab table extraction. Geometry-based row and column alignment relies on word positioning data to accurately reconstruct tables and apply confidence checks for clinical results.