Spaces:
Sleeping
Sleeping
File size: 2,152 Bytes
8fca640 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 | """Bounded native B-scan windows, including adjacent/distant and global views."""
import numpy as np
from src.io import iter_volume_frames, normalize_frame, read_frames
OFFSETS = {1: [0], 3: [-1, 0, 1], 5: [-8, -1, 0, 1, 8], 6: [-1, 0, 1]}
def iter_normalized_contexts(headers, shapes):
if any(shape[2] not in OFFSETS for shape in shapes):
raise ValueError("Unsupported OCT image context")
spatial = sorted({tuple(shape[:2]) for shape in shapes})
total = sum(int(ds.get("NumberOfFrames", 1)) for _, ds in headers)
if total < 1:
raise ValueError("Empty structural volume")
global_views = {}
if any(shape[2] == 6 for shape in shapes):
selected = [0, total // 2, total - 1]
raw = read_frames(headers, sorted(set(selected)))
for shape in spatial:
global_views[shape] = np.stack(
[normalize_frame(raw[index], shape) for index in selected]
)
radius = max(abs(offset) for shape in shapes for offset in OFFSETS[shape[2]])
iterator = iter(iter_volume_frames(headers))
cache = {}
loaded = -1
for center in range(total):
last = min(total - 1, center + radius)
while loaded < last:
frame = next(iterator, None)
if frame is None:
raise ValueError("Structural frames ended before declared geometry")
loaded += 1
cache[loaded] = {shape: normalize_frame(frame, shape) for shape in spatial}
images = []
for shape in shapes:
key = tuple(shape[:2])
indices = [
max(0, min(total - 1, center + offset)) for offset in OFFSETS[shape[2]]
]
image = np.stack([cache[index][key] for index in indices])
if shape[2] == 6:
image = np.concatenate([image, global_views[key]], axis=0)
images.append(image)
yield images
oldest = center + 1 - radius
for index in list(cache):
if index < oldest:
del cache[index]
if next(iterator, None) is not None:
raise ValueError("Surplus structural frames")
|