Upload 5 files

.gitattributes

@@ -33,3 +33,5 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+bin/cxr_ckd5_infer filter=lfs diff=lfs merge=lfs -text
+examples/sample_cxr.png filter=lfs diff=lfs merge=lfs -text

README.md

@@ -57,7 +57,7 @@ high-risk/low-risk stratification (Kaplan-Meier analyses).
 
 ## Intended use and limitations
 
-- **Research and screening-triage use only** — not a diagnostic device, not FDA
+- **Research use only** — not a diagnostic device, not FDA
   cleared/approved. Do not use to make individual clinical decisions.
 - Expects an **adult (≥18y), ambulatory, upright, PA chest radiograph**.
   Performance on portable/lateral views, pediatric patients, or CLAHE-processed
@@ -76,19 +76,10 @@ See the in-app disclaimer for more detail.
 - `app.py` — Gradio interface only (upload widget, results display, disclaimer).
   It does not load the CNN or the AFT model directly.
 - `bin/cxr_ckd5_infer` — a single compiled executable that bundles the CNN feature
-  extractor (Pyrros et al.) and the XGBoost AFT model together with their runtime
-  (Python, onnxruntime, xgboost). `app.py` invokes it as a subprocess, passing an
-  image path in and reading a JSON result back. The two model files are not present
-  in this repository as standalone, independently loadable files — only as
-  components embedded in this executable, consistent with the terms under which
-  the CNN weights may be redistributed.
+  extractor (Pyrros et al.) and the XGBoost AFT model together.
 - `examples/sample_cxr.png` — a public NIH ChestX-ray14 sample image, included only
   to demonstrate the interface (not part of the Emory/UIC study cohorts).
 
-The executable is built from `build/infer_cli.py` via PyInstaller inside a
-`python:3.10-slim-bullseye` container (see `build/Dockerfile.build`), which is not
-part of the deployed Space — it exists only to produce `bin/cxr_ckd5_infer`.
-
 ## License and attribution
 
 - The CNN feature extractor embedded in `bin/cxr_ckd5_infer` is © Ayis Pyrros et al.

bin/cxr_ckd5_infer

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8f33388521742287c0156376b6ba902c9a30d7ea2a30f2d7fcf5e2e505678045
+size 316179200