◉ SAP Field LabRESEARCH WORKSPACE
STANDARD AUTOMATED PERIMETRY

From report to research data.

Extract, review, and export your single field analysis.

01 Import report02 Review 52 loci03 Build your dataset24-2 · 52 loci · Two blind-spot loci excluded
+

Bulk import

Queue JPG, JPEG, PNG, WEBP, TIFF or TIF files together. Every TIFF page is processed separately. Each readable report is added as an unverified draft; use Review to check and correct it. Single-image import below works as before.

No files queued
File / pageStatusResult

Files and results remain only in this tab. Export your dataset before closing or refreshing. Pause finishes the current report before stopping.

01

Source report

JPG · JPEG · PNG · WEBP · TIFF · TIF
↥

Drop your SAP image here

Upload a clear, upright Humphrey 24-2 or 24-2C report.

Image processing happens in your browser.
Reports are held in this tab until you export or close it.
Ready for a report
Raw OCR text
02

Examination details

Awaiting image
Detected from the report title. Change this setting if needed.

Global indices

As printed on the report

Reliability indices

Missing ≠ zero

Threshold sensitivity dB

0 / 52 populated
Superior
Nasal (OD view)InferiorTemporal (OD view)
Missing OCR · review required Manually entered× Blind spot

Locus IDs use a common OD orientation. OS is mirrored on screen; exports include both visual-field and OD-normalized coordinates. Enter <0 exactly when printed.

Complete the required fields to verify this record.
03

Research dataset 0 examinations

Patient URPatient identifierTest dateEyeProgramMD (dB)PSD (dB)VFI (%)FPFNFixation lossLociReview
Your examinations will appear here. Export a wide table with the program and locus values, or a coordinate-based locus table.
WORKFLOW & DATA DICTIONARY

Traceable 24-2 and 24-2C datasets.

1. Upload a clear 24-2 or 24-2C single field report as JPG, JPEG, PNG, WEBP, TIFF, or TIF (up to 25 MB). Standard TIFFs are decoded locally, including common grayscale, RGB, LZW, Deflate, PackBits, JPEG, and CCITT scan formats. For a multipage TIFF, the first page opens automatically; use the TIFF page selector to process other pages. Add each examination before switching pages. BigTIFF and pages over 40 megapixels are unsupported. Header and index extraction starts automatically. Other programs (10-2, 30-2) and OCT reports are unsupported.

Bulk import: Choose multiple images in the Bulk import panel, then Start bulk import. All pages of a multipage TIFF become separate examinations. Reports are processed sequentially and readable results are saved as unverified drafts. Errors stay in the queue and do not stop other files. Pause stops after the active report; Resume continues pending reports. Review opens a saved draft with its original image; Save changes updates the same row. Clearing the queue keeps dataset rows, but releases their original images. Single-image import remains available below.

2. The app detects the eye and program. If the eye label is unreadable, an OD/OS token in the filename can suggest the eye and is explicitly flagged for confirmation. The reader detects the threshold chart axes, and number rows, then reads the 52 (24-2) or 62 (24-2C) loci automatically. If the chart is not detected, select the eye, drag around the numeric threshold plot, and click “Read selected plot”. Do not select grayscale, total deviation, or pattern deviation plots.

3. Check every field. Empty fields mean unavailable, never zero. FP/FN accept a percentage or printed fraction. Fixation loss preserves numerator/denominator. A printed “Off” or “Not tested” is retained as text, never converted to zero. Confirm both review boxes to mark a complete record verified. Draft records can be exported and are labelled unverified.

Coordinates: x increases toward the right of the printed visual-field plot; y increases superiorly. The 54-point 24-2 pattern uses 6° spacing. OD excludes (+15,+3) and (+15,−3); OS excludes (−15,+3) and (−15,−3). L01–L52 are ordered superior to inferior, left to right in OD orientation. OS locus IDs remain homologous after mirroring. 24-2C adds ten points (C01–C10), keeping L01–L52 stable. In OD visual-field coordinates, C01=(5,7), C02=(−1,5), C03=(1,5), C04=(7,5), C05=(−5,1), C06=(−7,−1), C07=(−7,−5), C08=(−1,−5), C09=(5,−7), C10=(1,−9), in degrees. The central detail includes the 12 standard loci within 10° and all ten added loci.

Exports: wide CSV has one row per examination, its test program and locus count, L01_db–L52_db, and C01_db–C10_db. C columns remain blank for 24-2 records. Printed MD, PSD, VFI and GHT are retained for the report’s program; they are not recalculated from a subset of loci. Locus CSV adds x/y, normalized x/y, source, confidence, and edit provenance. JSON includes original OCR text, selection, extraction timestamps, missing values and review status. Source images are not included. TIFF exports retain the original filename, source page number, and total page count. Leading spreadsheet formula characters are escaped in CSV.

Storage: patient reports stay in memory in this tab. Refreshing or closing clears the dataset. Export before leaving. OCR downloads its engine and English model from third-party CDNs; the image is processed locally, not sent to an OCR service.

Validation: this is an initial research tool. Tested against the supplied SAP sample and synthetic reports; broader validation is still needed. OCR can misread digits or place them in the wrong locus. It is not validated for clinical decisions.

ZEISS 24-2C pattern (page 4) · 52-location convention and blind-spot exclusions · Tesseract.js OCR

Excel CSV format: Every CSV includes a header row with units, including MD (dB), PSD (dB), VFI (%), False positive (%), False negative (%), and locus values (dB). Percentage cells contain plain numbers: 3% exports as 3; a reliability fraction of 1/20 exports as 5 in a percentage column. Fixation losses export as a decimal ratio from 0 to 1 (2/5 becomes 0.4; 20% becomes 0.2), avoiding Excel’s fraction-to-date conversion. Missing values stay blank and printed Off/Not tested values stay text. JSON retains the original printed values.