Datasets:

ClarusC64
/

ffr-physiological-plausibility-decay-detection-v0.1

id string	case_context string	ai_ffr_prediction float64	myocardial_perfusion_index float64	wall_motion_score float64	stress_test_result string	heart_rate int64	blood_pressure_systolic int64	physiological_coherence_score float64	expected_plausibility_band string	plausibility_gap float64	plausibility_decay_flag int64	decay_type string	modality_conflict_label string	plausibility_drop_score float64	notes string	constraints string	gold_checklist string
PPDD-001	normal physiology	0.84	0.88	1	normal	62	118	0.95	0.85-1.00	0.02	0	none	none	0.06	aligned	<=280 words	flag+type+conflict+drop
PPDD-002	mild perfusion defect consistent	0.8	0.75	1	mild defect	70	122	0.85	0.70-0.95	0.05	0	none	perfusion-consistent	0.18	within band	<=280 words	flag+type+conflict+drop
PPDD-003	perfusion normal but FFR severe	0.68	0.86	1	normal	64	120	0.55	0.80-1.00	0.22	1	perfusion-discordant	perfusion vs ffr	0.7	implausible ischemia call	<=280 words	flag+type+conflict+drop
PPDD-004	wall motion abnormal but FFR normal	0.84	0.62	0.55	abnormal	78	130	0.58	0.55-0.80	0.18	1	wall-motion-discordant	wall motion vs ffr	0.62	missed physiology	<=280 words	flag+type+conflict+drop
PPDD-005	multi-modality conflict rising	0.72	0.58	0.6	abnormal	85	138	0.48	0.55-0.75	0.16	1	multi-modality-conflict	perfusion + wall motion vs ffr	0.66	coherence breaking	<=280 words	flag+type+conflict+drop
PPDD-006	borderline mismatch	0.76	0.7	0.8	borderline	76	130	0.7	0.70-0.90	0.1	1	borderline-mismatch	soft conflict	0.44	watch and verify	<=280 words	flag+type+conflict+drop

Goal

Detect when an AI-derived FFR value
becomes physiologically implausible
given other modalities.

The warning signal is coherence loss.
Not a single bad threshold.

Inputs

Required outputs

Decay types

Examples:

Why cardiologists care

A stable-looking number
can still be wrong for the patient.

This dataset flags when:

Evaluation

The scorer checks that the response includes: