Spaces:
Sleeping
Sleeping
Download src/io.py from farrell236/OCT_Segmentation_a1: direct link, hf CLI and curl.
- Browser
- Download file 3.59 kB
-
https://huggingface.co/spaces/farrell236/OCT_Segmentation_a1/resolve/main/src/io.py
- Command line
-
hf download hf://spaces/farrell236/OCT_Segmentation_a1/src/io.py
-
curl -L -o io.py https://huggingface.co/spaces/farrell236/OCT_Segmentation_a1/resolve/main/src/io.py
3.59 kB
| """Label-free structural OCT I/O; no segmentation dependency.""" | |
| from pathlib import Path | |
| import json,re | |
| import numpy as np | |
| import pydicom | |
| from pydicom.pixels import pixel_array,iter_pixels | |
| from PIL import Image | |
| def load_manifest(path): | |
| rows=[json.loads(s) for s in Path(path).read_text().splitlines() if s.strip()] | |
| seen=set() | |
| for row in rows: | |
| name=row['case_id'] | |
| if not isinstance(name,str) or not re.fullmatch(r'[A-Za-z0-9_-]{1,64}',name):raise ValueError('Unsafe case identifier') | |
| if name in seen:raise ValueError('Duplicate case') | |
| seen.add(name) | |
| return rows | |
| def safe_path(root,relative): | |
| p=Path(relative) | |
| if p.is_absolute() or '..' in p.parts:raise ValueError('Input path must be dataset-relative') | |
| return Path(root)/p | |
| def volume_headers(root,oct_paths,with_spacing=False): | |
| tags=['SOPClassUID','Rows','Columns','NumberOfFrames','ConcatenationFrameOffsetNumber'] | |
| if with_spacing:tags+=['SharedFunctionalGroupsSequence','PerFrameFunctionalGroupsSequence'] | |
| # Inference asks for geometry, never participant identifiers or clinical tags. | |
| headers=[(safe_path(root,p),pydicom.dcmread(safe_path(root,p),stop_before_pixels=True,specific_tags=tags)) for p in oct_paths] | |
| headers.sort(key=lambda v:int(v[1].get('ConcatenationFrameOffsetNumber',0))) | |
| offsets=[int(d.get('ConcatenationFrameOffsetNumber',0)) for p,d in headers] | |
| n=0;h=int(headers[0][1].Rows);w=int(headers[0][1].Columns) | |
| for off,(p,d) in zip(offsets,headers): | |
| if off!=n:raise ValueError('Missing or overlapping concatenation frames') | |
| if (int(d.Rows),int(d.Columns))!=(h,w):raise ValueError('Inconsistent structural geometry') | |
| if str(d.SOPClassUID)!='1.2.840.10008.5.1.4.1.1.77.1.5.4':raise ValueError('Input is not structural OCT') | |
| n+=int(d.get('NumberOfFrames',1)) | |
| return headers,(n,h,w) | |
| def read_frame(headers,index): | |
| for path,ds in headers: | |
| off=int(ds.get('ConcatenationFrameOffsetNumber',0));n=int(ds.get('NumberOfFrames',1)) | |
| if off<=index<off+n: | |
| plugin='pylibjpeg' if ds.file_meta.TransferSyntaxUID.is_compressed else '' | |
| return pixel_array(path,index=index-off,decoding_plugin=plugin) | |
| raise IndexError('Frame outside volume') | |
| def read_frames(headers,indices): | |
| """Decode selected original frames with one file/decoder pass per source part.""" | |
| requested=[int(i) for i in indices];result={} | |
| for path,ds in headers: | |
| off=int(ds.get('ConcatenationFrameOffsetNumber',0));n=int(ds.get('NumberOfFrames',1)) | |
| local=[i-off for i in requested if off<=i<off+n] | |
| if not local:continue | |
| plugin='pylibjpeg' if ds.file_meta.TransferSyntaxUID.is_compressed else '' | |
| for i,frame in zip(local,iter_pixels(path,indices=local,decoding_plugin=plugin)): | |
| result[off+i]=frame | |
| if set(result)!=set(requested):raise ValueError('Selected structural frame coverage is incomplete') | |
| return result | |
| def iter_volume_frames(headers): | |
| for path,ds in headers: | |
| plugin='pylibjpeg' if ds.file_meta.TransferSyntaxUID.is_compressed else '' | |
| yield from iter_pixels(path,decoding_plugin=plugin) | |
| def normalize_frame(image,shape): | |
| """Deterministic within-image percentiles; no cohort fitting or metadata features.""" | |
| image=np.asarray(image,dtype=np.float32) | |
| finite=np.isfinite(image) | |
| if not finite.any():return np.zeros(shape,dtype=np.float32) | |
| lo,hi=np.quantile(image[finite],[.01,.995]) | |
| image=np.nan_to_num(image,nan=float(lo),posinf=float(hi),neginf=float(lo)) | |
| image=np.clip((image-lo)/max(float(hi-lo),1e-6),0,1) | |
| # PIL float bilinear matches center-based coordinate conversion. | |
| return np.asarray(Image.fromarray(image).resize((shape[1],shape[0]),Image.Resampling.BILINEAR),dtype=np.float32).copy() | |