Instructions to use Aditya2162/ivus-segmentation with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Keras
How to use Aditya2162/ivus-segmentation with Keras:
# Available backend options are: "jax", "torch", "tensorflow". import os os.environ["KERAS_BACKEND"] = "jax" import keras model = keras.saving.load_model("hf://Aditya2162/ivus-segmentation") - Notebooks
- Google Colab
- Kaggle
File size: 16,008 Bytes
1d197a4 | 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 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 | """Interactive GUI for editing per-frame IVUS contour annotations."""
from __future__ import annotations
import json
import os
from dataclasses import dataclass
from datetime import datetime
from glob import glob
from typing import Any
import matplotlib.pyplot as plt
import numpy as np
from matplotlib.lines import Line2D
from matplotlib.widgets import Button, Slider
from ..io.dicom import read_dicom
@dataclass
class FrameState:
"""In-memory annotation state for one frame."""
frame: int
lumen_x: list[float]
lumen_y: list[float]
plaque_x: list[float]
plaque_y: list[float]
lumen_confidence: float | None
plaque_confidence: float | None
bifurcation: bool = False
class AnnotationEditor:
"""Matplotlib-based editor for contour annotations."""
def __init__(
self,
dicom_path: str,
annotations_path: str,
edits_path: str,
) -> None:
self.dicom_path = dicom_path
self.annotations_path = annotations_path
self.edits_path = edits_path
_, images = read_dicom(dicom_path)
self.images = images
self.frame_states = self._load_base_annotations(annotations_path)
if len(self.frame_states) != images.shape[0]:
raise ValueError(
"Frame count mismatch between DICOM and annotations: "
f"{images.shape[0]} vs {len(self.frame_states)}"
)
self._load_existing_edits(edits_path)
self._ensure_edits_file_header()
self.current_frame = 0
self.drag_target: tuple[str, int] | None = None
self.drag_threshold_px = 10.0
self.frame_dirty = False
self.fig = None
self.ax_image = None
self.image_artist = None
self.lumen_line: Line2D | None = None
self.plaque_line: Line2D | None = None
self.lumen_points = None
self.plaque_points = None
self.slider: Slider | None = None
self.status_text = None
self.is_updating_slider = False
@staticmethod
def _safe_float(value: Any) -> float | None:
if value is None:
return None
try:
out = float(value)
except (TypeError, ValueError):
return None
if np.isnan(out):
return None
return out
def _load_base_annotations(self, path: str) -> list[FrameState]:
states: list[FrameState] = []
with open(path, "r", encoding="utf-8") as fp:
for raw in fp:
line = raw.strip()
if not line:
continue
rec = json.loads(line)
if rec.get("record_type") != "frame":
continue
states.append(
FrameState(
frame=int(rec["frame"]),
lumen_x=[float(v) for v in rec.get("lumen", {}).get("x", [])],
lumen_y=[float(v) for v in rec.get("lumen", {}).get("y", [])],
plaque_x=[float(v) for v in rec.get("plaque", {}).get("x", [])],
plaque_y=[float(v) for v in rec.get("plaque", {}).get("y", [])],
lumen_confidence=self._safe_float(rec.get("lumen_confidence")),
plaque_confidence=self._safe_float(rec.get("plaque_confidence")),
bifurcation=bool(rec.get("bifurcation", False)),
)
)
states.sort(key=lambda s: s.frame)
expected = list(range(len(states)))
actual = [s.frame for s in states]
if actual != expected:
raise ValueError(
"Base annotations must contain contiguous frame records in order "
f"(got first frames: {actual[:10]})"
)
return states
def _load_existing_edits(self, path: str) -> None:
if not os.path.exists(path):
return
latest: dict[int, dict[str, Any]] = {}
with open(path, "r", encoding="utf-8") as fp:
for raw in fp:
line = raw.strip()
if not line:
continue
rec = json.loads(line)
if rec.get("record_type") != "frame_edit":
continue
frame_idx = int(rec["frame"])
latest[frame_idx] = rec
for frame_idx, rec in latest.items():
if frame_idx < 0 or frame_idx >= len(self.frame_states):
continue
state = self.frame_states[frame_idx]
state.lumen_x = [float(v) for v in rec.get("lumen", {}).get("x", state.lumen_x)]
state.lumen_y = [float(v) for v in rec.get("lumen", {}).get("y", state.lumen_y)]
state.plaque_x = [float(v) for v in rec.get("plaque", {}).get("x", state.plaque_x)]
state.plaque_y = [float(v) for v in rec.get("plaque", {}).get("y", state.plaque_y)]
state.bifurcation = bool(rec.get("bifurcation", state.bifurcation))
def _ensure_edits_file_header(self) -> None:
os.makedirs(os.path.dirname(os.path.abspath(self.edits_path)), exist_ok=True)
if os.path.exists(self.edits_path):
return
meta = {
"record_type": "meta",
"created_at": datetime.now().isoformat(timespec="seconds"),
"dicom_path": os.path.abspath(self.dicom_path),
"base_annotations_path": os.path.abspath(self.annotations_path),
"format": "append_only_frame_edits",
}
with open(self.edits_path, "w", encoding="utf-8") as fp:
fp.write(json.dumps(meta) + "\n")
def _append_frame_edit(self, frame_idx: int, reason: str) -> None:
state = self.frame_states[frame_idx]
rec = {
"record_type": "frame_edit",
"saved_at": datetime.now().isoformat(timespec="seconds"),
"reason": reason,
"frame": frame_idx,
"bifurcation": state.bifurcation,
"lumen": {"x": state.lumen_x, "y": state.lumen_y},
"plaque": {"x": state.plaque_x, "y": state.plaque_y},
"lumen_confidence": state.lumen_confidence,
"plaque_confidence": state.plaque_confidence,
}
with open(self.edits_path, "a", encoding="utf-8") as fp:
fp.write(json.dumps(rec) + "\n")
fp.flush()
os.fsync(fp.fileno())
self.frame_dirty = False
def _build_ui(self) -> None:
self.fig = plt.figure(figsize=(13, 9))
self.fig.canvas.manager.set_window_title("DeepIVUS Annotation Editor")
self.ax_image = self.fig.add_axes([0.05, 0.22, 0.9, 0.74])
self.ax_image.set_title("DeepIVUS Annotation Editor", fontsize=14, weight="bold")
self.ax_image.set_axis_off()
slider_ax = self.fig.add_axes([0.12, 0.13, 0.76, 0.035])
self.slider = Slider(
ax=slider_ax,
label="Frame",
valmin=0,
valmax=len(self.frame_states) - 1,
valinit=0,
valstep=1,
color="#1f77b4",
)
prev_ax = self.fig.add_axes([0.12, 0.05, 0.1, 0.055])
next_ax = self.fig.add_axes([0.24, 0.05, 0.1, 0.055])
save_ax = self.fig.add_axes([0.42, 0.05, 0.14, 0.055])
bif_ax = self.fig.add_axes([0.72, 0.05, 0.16, 0.055])
self.prev_button = Button(prev_ax, "Prev Frame", color="#E0E0E0", hovercolor="#D0D0D0")
self.next_button = Button(next_ax, "Next Frame", color="#E0E0E0", hovercolor="#D0D0D0")
self.save_button = Button(save_ax, "Save Frame", color="#D6F5D6", hovercolor="#BFF0BF")
self.bif_button = Button(bif_ax, "Bifurcation: No", color="#F5D6D6", hovercolor="#F0BFBF")
self.status_text = self.fig.text(0.05, 0.18, "", fontsize=10)
self.slider.on_changed(self._on_slider)
self.prev_button.on_clicked(self._on_prev)
self.next_button.on_clicked(self._on_next)
self.save_button.on_clicked(self._on_save)
self.bif_button.on_clicked(self._on_toggle_bifurcation)
self.fig.canvas.mpl_connect("button_press_event", self._on_press)
self.fig.canvas.mpl_connect("motion_notify_event", self._on_motion)
self.fig.canvas.mpl_connect("button_release_event", self._on_release)
self.fig.canvas.mpl_connect("key_press_event", self._on_key)
self.fig.canvas.mpl_connect("close_event", self._on_close)
def _state(self) -> FrameState:
return self.frame_states[self.current_frame]
def _render_frame(self) -> None:
state = self._state()
image = self.images[self.current_frame]
if self.image_artist is None:
self.image_artist = self.ax_image.imshow(image, cmap="gray")
self.lumen_line = self.ax_image.plot([], [], color="#1db954", lw=2)[0]
self.plaque_line = self.ax_image.plot([], [], color="#ff5a5a", lw=2)[0]
self.lumen_points = self.ax_image.scatter([], [], c="#1db954", s=28, edgecolors="black", linewidths=0.4)
self.plaque_points = self.ax_image.scatter([], [], c="#ff5a5a", s=28, edgecolors="black", linewidths=0.4)
else:
self.image_artist.set_data(image)
lumen_x, lumen_y = state.lumen_x, state.lumen_y
plaque_x, plaque_y = state.plaque_x, state.plaque_y
self.lumen_line.set_data(lumen_x + lumen_x[:1], lumen_y + lumen_y[:1])
self.plaque_line.set_data(plaque_x + plaque_x[:1], plaque_y + plaque_y[:1])
lumen_offsets = np.c_[lumen_x, lumen_y] if lumen_x and lumen_y else np.empty((0, 2))
plaque_offsets = np.c_[plaque_x, plaque_y] if plaque_x and plaque_y else np.empty((0, 2))
self.lumen_points.set_offsets(lumen_offsets)
self.plaque_points.set_offsets(plaque_offsets)
bif_text = "Yes" if state.bifurcation else "No"
bif_color = "#D6F5D6" if state.bifurcation else "#F5D6D6"
self.bif_button.label.set_text(f"Bifurcation: {bif_text}")
self.bif_button.ax.set_facecolor(bif_color)
self.status_text.set_text(
f"Frame {self.current_frame + 1}/{len(self.frame_states)} "
f"Lumen pts: {len(lumen_x)} Plaque pts: {len(plaque_x)} "
f"Autosave file: {os.path.basename(self.edits_path)}"
)
self.fig.canvas.draw_idle()
def _set_frame(self, frame_idx: int) -> None:
frame_idx = int(np.clip(frame_idx, 0, len(self.frame_states) - 1))
if frame_idx == self.current_frame:
return
if self.frame_dirty:
self._append_frame_edit(self.current_frame, reason="frame_change")
self.current_frame = frame_idx
self.is_updating_slider = True
self.slider.set_val(frame_idx)
self.is_updating_slider = False
self._render_frame()
def _nearest_point(self, x: float, y: float) -> tuple[str, int] | None:
state = self._state()
def best_idx(xs: list[float], ys: list[float]) -> tuple[int, float] | None:
if not xs:
return None
pts = np.column_stack((np.asarray(xs), np.asarray(ys)))
dist = np.linalg.norm(pts - np.asarray([x, y]), axis=1)
idx = int(np.argmin(dist))
return idx, float(dist[idx])
lumen = best_idx(state.lumen_x, state.lumen_y)
plaque = best_idx(state.plaque_x, state.plaque_y)
choice: tuple[str, int, float] | None = None
if lumen is not None:
choice = ("lumen", lumen[0], lumen[1])
if plaque is not None and (choice is None or plaque[1] < choice[2]):
choice = ("plaque", plaque[0], plaque[1])
if choice is None or choice[2] > self.drag_threshold_px:
return None
return choice[0], choice[1]
def _on_slider(self, val: float) -> None:
if self.is_updating_slider:
return
self._set_frame(int(val))
def _on_prev(self, _event: Any) -> None:
self._set_frame(self.current_frame - 1)
def _on_next(self, _event: Any) -> None:
self._set_frame(self.current_frame + 1)
def _on_save(self, _event: Any) -> None:
self._append_frame_edit(self.current_frame, reason="manual_save")
self._render_frame()
def _on_toggle_bifurcation(self, _event: Any) -> None:
state = self._state()
state.bifurcation = not state.bifurcation
self.frame_dirty = True
self._append_frame_edit(self.current_frame, reason="bifurcation_toggle")
self._render_frame()
def _on_key(self, event: Any) -> None:
if event.key in {"left", "a"}:
self._set_frame(self.current_frame - 1)
elif event.key in {"right", "d"}:
self._set_frame(self.current_frame + 1)
elif event.key == "s":
self._on_save(event)
elif event.key == "b":
self._on_toggle_bifurcation(event)
def _on_press(self, event: Any) -> None:
if event.inaxes != self.ax_image or event.button != 1:
return
if event.xdata is None or event.ydata is None:
return
hit = self._nearest_point(float(event.xdata), float(event.ydata))
self.drag_target = hit
def _on_motion(self, event: Any) -> None:
if self.drag_target is None:
return
if event.inaxes != self.ax_image or event.xdata is None or event.ydata is None:
return
state = self._state()
contour_name, idx = self.drag_target
h, w = self.images[self.current_frame].shape
x = float(np.clip(event.xdata, 0, w - 1))
y = float(np.clip(event.ydata, 0, h - 1))
if contour_name == "lumen":
state.lumen_x[idx] = x
state.lumen_y[idx] = y
else:
state.plaque_x[idx] = x
state.plaque_y[idx] = y
self.frame_dirty = True
self._render_frame()
def _on_release(self, _event: Any) -> None:
if self.drag_target is None:
return
self.drag_target = None
if self.frame_dirty:
self._append_frame_edit(self.current_frame, reason="point_drag")
self._render_frame()
def _on_close(self, _event: Any) -> None:
if self.frame_dirty:
self._append_frame_edit(self.current_frame, reason="window_close")
def show(self) -> None:
self._build_ui()
self._render_frame()
plt.show()
def _default_annotations_path(dicom_path: str, output_root: str) -> str:
stem = os.path.splitext(os.path.basename(dicom_path))[0]
pattern = os.path.join(output_root, "*", f"{stem}_contours.jsonl")
matches = [p for p in glob(pattern) if os.path.isfile(p)]
if matches:
matches.sort(key=lambda p: os.path.getmtime(p), reverse=True)
return matches[0]
raise FileNotFoundError(
"No pipeline contour JSONL found for "
f"'{stem}'. Expected files like: {pattern}"
)
def _default_edits_path(annotations_path: str) -> str:
if annotations_path.endswith("_contours.jsonl"):
return annotations_path.replace("_contours.jsonl", "_edited_annotations.jsonl")
root, _ = os.path.splitext(annotations_path)
return root + "_edited_annotations.jsonl"
def launch_annotation_editor(
dicom_path: str,
annotations_path: str | None = None,
edits_path: str | None = None,
output_root: str = "output",
) -> None:
"""Launch interactive annotation GUI.
If ``annotations_path`` is not provided, the latest contours JSONL is
selected from ``output/<timestamp>/``.
"""
if annotations_path is None:
annotations_path = _default_annotations_path(dicom_path, output_root)
if edits_path is None:
edits_path = _default_edits_path(annotations_path)
editor = AnnotationEditor(
dicom_path=dicom_path,
annotations_path=annotations_path,
edits_path=edits_path,
)
editor.show()
|