[dnncha]

Why

compare_segmentation() converts the default minRR=None to infinity. After the first match, every unmatched detection then enters the short-RR suppression path and is omitted from the false-positive count. Keep None as no filter and apply suppression only when the caller provides minRR.

Fixes #85.

Scope

• Count unmatched detections as false positives when minRR is omitted.
• Preserve the existing short-RR filter for explicit minRR values.
• Add focused coverage for both paths.

Blast Radius

This changes the default false-positive count, acc, and err from compare_segmentation(). True-positive counts and sensitivity remain unchanged, and explicit minRR behavior remains available. The public-data replay establishes changed benchmark values on one ECG record, but no affected publication or clinical conclusion was identified. This correction has not received independent human scientific review.

Verification

• .venv/bin/python -m pytest -q tests/test_ecg_compare_segmentation.py passed both focused tests.
• On ten minutes of MIT-BIH record 108, false-positive counts from five BioSPPy segmenters matched wfdb.processing.compare_annotations after the fix.
• The perfect-match and explicit-minRR controls retained TP=3 and FP=0.
• .venv/bin/python -m compileall -q biosppy tests passed.