Pipeline for generating report for ACE-D Aim 3 randomization.

Setup

Local:

pip install -r requirements.txt

Docker (no local Python needed):

docker build -t rand-pipeline .

Usage

Local:

python pipeline.py <wn_csv> <ec_csv> <output.xlsx> [--subset-wn-tests]

--subset-wn-tests restricts the WebNeuro tab/chart to the tests used by the composite scores plus Switching of Attention 1 (see below); omit it to keep every WebNeuro test.

Docker — mount your current directory to /workspace so input/output files are visible on both sides:

macOS/Linux (bash/zsh):

docker run --rm -v "$(pwd):/workspace" rand-pipeline python pipeline.py <wn_csv> <ec_csv> <output.xlsx>

Windows PowerShell:

docker run --rm -v "${PWD}:/workspace" rand-pipeline python pipeline.py <wn_csv> <ec_csv> <output.xlsx>

Windows cmd.exe:

docker run --rm -v "%cd%:/workspace" rand-pipeline python pipeline.py <wn_csv> <ec_csv> <output.xlsx>

Note: on Windows, use PowerShell or cmd.exe for this command, not Git Bash — Git Bash's $(pwd) produces a POSIX-style path that Docker Desktop on Windows doesn't translate correctly, so the mount silently points to the wrong location.

Reads two CSVs (tolerating ragged rows — stray trailing commas, rows shorter or longer than the header), strips whitespace/quoting from headers and values, converts numeric-looking columns, and writes each CSV to its own tab in a single Excel workbook (wn_csv → "WebNeuro" tab, ec_csv → "EtCere" tab, fixed).

The WebNeuro tab's columns are reordered: identifying/demographic info (ID, Session, Age, Gender, TestDate) first, then raw test variables, then normed (_norm) test variables — each group ordered by the WebNeuro test battery's administration order. The variable order is hardcoded in pipeline.py (WN_RAW_VARIABLE_ORDER).

With --subset-wn-tests, only the tests used by the composite scores, plus Switching of Attention 1, are kept — Digit Span (Forward), Stroop Word, Stroop Color, Switching of Attention 1/2, GoNo-Go, and Maze (WN_REPORTED_TESTS in pipeline.py). Every other WebNeuro test's raw/normed variables are dropped from the tab and chart.

The first two rows of wn_csv are kept (a WebNeuro export can contain one row per session; the WebNeuro tab reflects the first two sessions — screening and baseline).

The WebNeuro tab also gets 6 composite score columns appended at the end — maze_composite, gng_composite, stroopw_composite, stroopc_composite, swoa_composite, and digit_composite — each the row-wise mean of a subset of that test's normed variables (WN_COMPOSITE_GROUPS in pipeline.py).

Only two rows of ec_csv are kept: the "Referenced " row (row 10, EC_REFERENCED_ROW in pipeline.py — its label includes the participant's name, so it's picked by row position rather than a label match) and the QC metric rows (Signal-to-Noise Ratio, Critical Motion Control — matched by label prefix, EC_QC_LABEL_PREFIXES). The rest of the EtCere export's fixed report template (raw/global scores, healthy-norm stats, etc.) is dropped.

The workbook also gets a "Norm Score Chart" tab: a line chart of the normed scores for both sessions, one dashed line per session (WN_SESSION_DASH_STYLES in pipeline.py; both lines are the same color and told apart by dash pattern instead), labeled "Screening" and "Baseline" (WN_SESSION_LABELS in pipeline.py). Each variable's column is shaded with a background band colored by the WebNeuro test it belongs to (WN_TEST_GROUPS). The y-axis has value ticks, and the x-axis shows every variable name (rotated 45° to fit all 64). The legend (bottom) lists the two session lines.

Test data

data/ is git-ignored and excluded from the built Docker image (see .dockerignore) — it's only meant to hold local participant CSVs for manual testing, never to be committed or shipped in an image.