Commit History
-
f
convitomc9b4129
-
f
convitom6a13626 -
f
convitom320063f
-
f
convitomd313d81
-
f
convitom8356dae
-
f
convitomb961b41
-
feat(chexpert): U-MultiClass PNU abnormality guidance + abnormality-guided VQA
convitom +1: Claude Sonnet 4.6- chexpert_classifier.py: 14 binary heads → 14×3 softmax (negative/ positive/uncertain per pathology, META-CXR / CheXpert U-MultiClass). Add format_pnu/buckets_to_pnu as the single source of truth for the PNU 3-section prompt string (shared with the oracle builder so GT and predicted prompts are byte-identical). - mimic_cxr_builder.py: GT chexpert.csv → PNU string (1→pos, 0→neg, -1→uncertain, blank/NaN→neg). VQA now carries the SAME PNU context (abnormality-guided VQA, RaDialog-style). O(1) image lookup (was O(N²)); index every image so report-less studies still serve VQA. - Drop the obsolete uncertain_policy knob (U-MultiClass is the only behaviour now) from builder, dataset.py, dataset_resolver, config. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>215ecd6
-
feat(data): add split_cascade report mode + MIMIC-CXR builder with CheXpert oracle labels
convitom +1: Claude Sonnet 4.6- New report_mode "split_cascade": findings task unchanged (image+CheXpert), impression task takes GT findings text as prompt context (findings→impression summarisation). Wired through iu_xray_builder, dataset_resolver, config docs. - New data/mimic_cxr_builder.py: parses the pre-split MIMIC layout, bakes the 14 GT CheXpert labels (oracle, from *chexpert*.csv) into structured_findings as "Predicted Findings: ..." — the RaDialog image + abnormality-guidance setup. Supports all 3 report_mode / image_mode axes; optional VQA attach. - Implement build_instruct_json() (was NotImplementedError) as a thin delegate. - dataset_resolver: auto-build MIMIC JSON (mode-suffixed cache) when missing. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>02426e6
-
initial commit
convitom28b13fc