File size: 10,803 Bytes
f12fd7f |
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 |
"""
Gradio app to explore pancreas or lymphome clinical report annotations.
"""
import os
from functools import partial
from pathlib import Path
import gradio as gr
from datasets import load_dataset
MIN_ANNOTATIONS = 10
PANCREAS_REPO_ID = os.getenv("PANCREAS_REPO_ID", "rntc/biomed-fr-pancreas-annotations")
LYMPHOME_REPO_ID = os.getenv("LYMPHOME_REPO_ID", "rntc/biomed-fr-lymphome-annotations")
LYMPHOME_LOCAL_JSONL = (
Path(__file__).resolve().parent.parent
/ "Qwen--Qwen3-235B-A22B-Instruct-2507-FP8-4-lymphome-annotation-20251201_153807.jsonl"
)
# Colors for highlighting
COLORS = [
"#FFEB3B",
"#4CAF50",
"#2196F3",
"#FF9800",
"#E91E63",
"#9C27B0",
"#00BCD4",
"#8BC34A",
"#FF5722",
"#607D8B",
]
def count_real_annotations(annotation):
"""Count real annotations (excluding 'not found' placeholders)."""
count = 0
for var_data in annotation.values():
if var_data and isinstance(var_data, dict):
value = var_data.get("value")
span = var_data.get("span", "")
if value:
if span and "pas de mention" in span.lower():
continue
if "not performed" in str(value).lower():
continue
count += 1
return count
def escape_html(text):
if not text:
return ""
return str(text).replace("&", "&").replace("<", "<").replace(">", ">")
def highlight_text(cr_text, annotation):
"""Highlight spans in CR text."""
if not cr_text or not annotation:
return f"<pre style='white-space:pre-wrap;'>{escape_html(cr_text)}</pre>"
# Collect valid spans (that exist in text)
spans = []
for var_name, var_data in annotation.items():
if var_data and isinstance(var_data, dict):
span = var_data.get("span")
value = var_data.get("value")
if span and value and span in cr_text:
spans.append(
{
"text": span,
"start": cr_text.find(span),
"var": var_name.replace("_", " ").title(),
"value": str(value),
}
)
if not spans:
return f"<pre style='white-space:pre-wrap;'>{escape_html(cr_text)}</pre>"
# Sort by position and remove overlaps
spans.sort(key=lambda x: x["start"])
filtered = []
for s in spans:
s["end"] = s["start"] + len(s["text"])
if not filtered or s["start"] >= filtered[-1]["end"]:
filtered.append(s)
# Build HTML
html = []
pos = 0
color_map = {}
color_idx = 0
for s in filtered:
if s["start"] > pos:
html.append(escape_html(cr_text[pos : s["start"]]))
if s["var"] not in color_map:
color_map[s["var"]] = COLORS[color_idx % len(COLORS)]
color_idx += 1
color = color_map[s["var"]]
html.append(
f'<mark style="background:{color};padding:1px 3px;border-radius:3px;" '
f'title="{escape_html(s["var"])}: {escape_html(s["value"])}">'
f'{escape_html(s["text"])}</mark>'
)
pos = s["end"]
if pos < len(cr_text):
html.append(escape_html(cr_text[pos:]))
return f"<pre style='white-space:pre-wrap;line-height:1.6;'>{''.join(html)}</pre>"
def format_table(annotation):
"""Format annotations as HTML table."""
if not annotation:
return "<p>No annotations</p>"
rows = []
for var_name, var_data in annotation.items():
if var_data and isinstance(var_data, dict):
value = var_data.get("value")
span = var_data.get("span", "")
var_label = var_name.replace("_", " ").title()
if value:
if span and "pas de mention" in span.lower():
display_value = "/"
display_span = ""
elif "not performed" in str(value).lower():
display_value = "/"
display_span = ""
else:
display_value = str(value)
display_span = span[:60] + "..." if span and len(span) > 60 else (span or "")
else:
display_value = "/"
display_span = ""
rows.append(
f"""<tr>
<td style="padding:6px 10px;border-bottom:1px solid #ddd;font-weight:500;">{escape_html(var_label)}</td>
<td style="padding:6px 10px;border-bottom:1px solid #ddd;color:#1565C0;">{escape_html(display_value)}</td>
<td style="padding:6px 10px;border-bottom:1px solid #ddd;color:#666;font-size:12px;font-style:italic;">{escape_html(display_span)}</td>
</tr>"""
)
return f"""<table style="width:100%;border-collapse:collapse;font-size:13px;">
<thead><tr style="background:#f5f5f5;">
<th style="padding:8px 10px;text-align:left;border-bottom:2px solid #ddd;">Variable</th>
<th style="padding:8px 10px;text-align:left;border-bottom:2px solid #ddd;">Value</th>
<th style="padding:8px 10px;text-align:left;border-bottom:2px solid #ddd;">Source</th>
</tr></thead>
<tbody>{"".join(rows)}</tbody>
</table>"""
def load_pancreas_dataset():
print(f"Loading pancreas dataset from {PANCREAS_REPO_ID}...")
dataset = load_dataset(PANCREAS_REPO_ID, split="train")
print(f"Loaded {len(dataset)} pancreas samples")
return dataset
def load_lymphome_dataset():
print(f"Loading lymphome dataset from {LYMPHOME_REPO_ID} (Hub)...")
try:
dataset = load_dataset(LYMPHOME_REPO_ID, split="train")
print(f"Loaded {len(dataset)} lymphome samples from Hub")
return dataset
except Exception as exc: # noqa: BLE001 (we want to surface any failure)
print(f"Failed to load lymphome dataset from Hub: {exc}")
if LYMPHOME_LOCAL_JSONL.exists():
print(f"Falling back to local lymphome JSONL at {LYMPHOME_LOCAL_JSONL}")
dataset = load_dataset("json", data_files=str(LYMPHOME_LOCAL_JSONL), split="train")
print(f"Loaded {len(dataset)} lymphome samples from local file")
return dataset
raise
def filter_indices(dataset, min_annotations):
return [
i
for i, sample in enumerate(dataset)
if count_real_annotations(sample.get("annotation", {})) >= min_annotations
]
def prepare_source(key, label, loader, min_annotations):
"""Load a dataset source and precompute filtered indices."""
try:
dataset = loader()
filtered = filter_indices(dataset, min_annotations)
print(f"{label}: filtered to {len(filtered)} samples with >= {min_annotations} annotations")
return {
"label": label,
"dataset": dataset,
"filtered_indices": filtered,
"min_annotations": min_annotations,
"error": None,
}
except Exception as exc: # noqa: BLE001 (we want to surface any failure)
print(f"Failed to load {label}: {exc}")
return {
"label": label,
"dataset": None,
"filtered_indices": [],
"min_annotations": min_annotations,
"error": str(exc),
}
SOURCES = {
"pancreas": prepare_source("pancreas", "Pancréas", load_pancreas_dataset, MIN_ANNOTATIONS),
"lymphome": prepare_source("lymphome", "Lymphome", load_lymphome_dataset, MIN_ANNOTATIONS),
}
def display_sample_for_source(source_key, slider_idx):
"""Display a sample for a given dataset source."""
source = SOURCES[source_key]
if source["error"]:
message = f"Dataset unavailable: {source['error']}"
return message, message, message
if not source["filtered_indices"]:
message = f"No samples with >= {source['min_annotations']} annotations."
return message, message, message
slider_idx = int(slider_idx)
if slider_idx < 0 or slider_idx >= len(source["filtered_indices"]):
return "Invalid", "Invalid", "Invalid"
real_idx = source["filtered_indices"][slider_idx]
sample = source["dataset"][real_idx]
original = sample.get("original_text", "")
cr = sample.get("CR", "")
annotation = sample.get("annotation", {})
n_annotations = count_real_annotations(annotation)
original_html = f"<pre style='white-space:pre-wrap;line-height:1.6;'>{escape_html(original)}</pre>"
cr_html = (
f"<p><b>Sample #{real_idx}</b> — {n_annotations} annotations</p>"
+ highlight_text(cr, annotation)
)
return original_html, cr_html, format_table(annotation)
def build_tab(source_key):
source = SOURCES[source_key]
label = source["label"]
with gr.TabItem(label):
if source["error"]:
gr.Markdown(f"⚠️ Could not load {label} dataset: {escape_html(source['error'])}")
return
if not source["filtered_indices"]:
gr.Markdown(f"⚠️ No samples with >= {source['min_annotations']} annotations.")
return
gr.Markdown(
f"Showing {len(source['filtered_indices'])} samples with >= "
f"{source['min_annotations']} annotations. Hover over highlights to see values."
)
with gr.Row():
slider = gr.Slider(
0,
len(source["filtered_indices"]) - 1,
value=0,
step=1,
label="Sample",
)
with gr.Row():
with gr.Column():
gr.Markdown("### Original (English)")
original_html = gr.HTML()
with gr.Column():
gr.Markdown("### Generated CR (French)")
cr_html = gr.HTML()
with gr.Column():
gr.Markdown("### Extracted Variables")
table_html = gr.HTML()
slider.change(
fn=partial(display_sample_for_source, source_key),
inputs=[slider],
outputs=[original_html, cr_html, table_html],
)
demo.load(
fn=partial(display_sample_for_source, source_key),
inputs=[slider],
outputs=[original_html, cr_html, table_html],
)
# Build UI
with gr.Blocks(title="Clinical Annotations Explorer", theme=gr.themes.Base()) as demo:
gr.Markdown("# 🔬 Clinical Annotation Explorer")
gr.Markdown(
"Use the tabs below to switch between pancreas and lymphome annotations. "
"Hover over highlights to see the extracted values."
)
with gr.Tabs():
build_tab("pancreas")
build_tab("lymphome")
if __name__ == "__main__":
demo.launch()
|