-
Notifications
You must be signed in to change notification settings - Fork 1
Expand file tree
/
Copy pathindex.html
More file actions
675 lines (632 loc) · 27.2 KB
/
index.html
File metadata and controls
675 lines (632 loc) · 27.2 KB
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
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
<!DOCTYPE html>
<html>
<head>
<meta charset="utf-8">
<!--Hepato-LLaVA-->
<meta name="description"
content="Hepato-LLaVA: An Expert MLLM with Sparse Topo-Pack Attention for Hepatocellular Pathology Analysis on Whole Slide Images">
<meta name="keywords" content="Hepato-LLaVA, Hepatocellular Carcinoma, Multi-modal Large Language Models, Whole Slide Images, Digital Pathology">
<meta name="viewport" content="width=device-width, initial-scale=1">
<title>Hepato-LLaVA</title>
<link rel="icon" href="./static/images/icon.png">
<script async src="https://www.googletagmanager.com/gtag/js?id=G-PYVRSFMDRL"></script>
<script>
window.dataLayer = window.dataLayer || [];
function gtag() {
dataLayer.push(arguments);
}
gtag('js', new Date());
gtag('config', 'G-PYVRSFMDRL');
</script>
<link href="https://fonts.googleapis.com/css?family=Google+Sans|Noto+Sans|Castoro" rel="stylesheet">
<!--css file-->
<link rel="stylesheet" href="./static/css/bulma.min.css">
<link rel="stylesheet" href="./static/css/bulma-carousel.min.css">
<link rel="stylesheet" href="./static/css/bulma-slider.min.css">
<link rel="stylesheet" href="./static/css/fontawesome.all.min.css">
<link rel="stylesheet" href="https://cdn.jsdelivr.net/gh/jpswalsh/academicons@1/css/academicons.min.css">
<link rel="stylesheet" href="./static/css/index.css">
<link rel="stylesheet" href="./static/css/leaderboard.css">
<!--html head icon-->
<link rel="icon" href="./static/images/icon.png">
<!--js file-->
<script type="text/javascript" src="./static/js/sort-table.js" defer></script>
<script src="https://ajax.googleapis.com/ajax/libs/jquery/3.5.1/jquery.min.js"></script>
<script defer src="./static/js/fontawesome.all.min.js"></script>
<script src="./static/js/bulma-carousel.min.js"></script>
<script src="./static/js/bulma-slider.min.js"></script>
<script src="./static/js/index.js"></script>
<script src="./static/js/leaderboard_button.js"></script>
</head>
<!--Start Body-->
<body>
<!--Nav Content-->
<nav class="navbar" role="navigation" aria-label="main navigation">
<div class="navbar-brand">
<a role="button" class="navbar-burger" aria-label="menu" aria-expanded="false">
<span aria-hidden="true"></span>
<span aria-hidden="true"></span>
<span aria-hidden="true"></span>
</a>
</div>
<!--nav menu-->
<div class="navbar-menu">
<div class="navbar-start" style="flex-grow: 1; justify-content: center;">
<!--navbar link-->
<div class="navbar-item has-dropdown is-hoverable">
<a class="navbar-link">
More Research
</a>
<div class="navbar-dropdown">
<a class="navbar-item">
Future Release
</a>
</div>
</div>
</div>
</div>
</nav>
<!--title author and link-->
<section class="hero">
<div class="hero-body">
<div class="container is-max-desktop">
<div class="columns is-centered">
<div class="column has-text-centered">
<!--main title-->
<h1 class="title is-1 publication-title">
<img src="static/images/icon.png" style="width:2.5em; vertical-align: middle" alt="Logo"/>
<span class="hepato-brand" style="vertical-align: middle">Hepato-LLaVA</span>
</h1>
<!--subtitle-->
<h2 class="subtitle is-3 publication-subtitle">
Hepato-LLaVA: An Expert MLLM with Sparse Topo-Pack Attention for Hepatocellular Pathology Analysis on Whole Slide Images
</h2>
<!--author link-->
<div class="is-size-5 publication-authors">
<span class="author-block">
Yuxuan Yang*<sup style="color:#4B70F5;"><b>1</b></sup>,
</span>
<span class="author-block">
Zhonghao Yan*†<sup style="color:#4B70F5;"><b>1</b></sup>,
</span>
<span class="author-block">
Yi Zhang*<sup style="color:#06D001;"><b>2</b></sup>,
</span>
<span class="author-block">
Bo Yun<sup style="color:#4B70F5;"><b>1</b></sup>,
</span>
<span class="author-block">
Muxi Diao<sup style="color:#4B70F5;"><b>1</b></sup>,
</span>
<span class="author-block">
Guowei Zhao<sup style="color:#06D001;"><b>2</b></sup>,
</span>
<span class="author-block">
Kongming Liang‡<sup style="color:#4B70F5;"><b>1</b></sup>,
</span>
<span class="author-block">
Wenbin Li‡<sup style="color:#06D001;"><b>2</b></sup>,
</span>
<span class="author-block">
Zhanyu Ma<sup style="color:#4B70F5;"><b>1</b></sup>
</span>
</div>
<!--author institution-->
<div class="is-size-6 publication-authors" style="margin-top: 0.5em;">
<span class="author-block"><sup style="color:#4B70F5;"><b>1</b></sup>Department of Artificial Intelligence, Beijing University of Posts and Telecommunications, Beijing 100876, China</span>
<span class="author-block"><sup style="color:#06D001;"><b>2</b></sup>Department of Pathology, National Cancer Center/National Clinical Research Center for Cancer/Cancer Hospital, Chinese Academy of Medical Sciences and Peking Union Medical College, Beijing 100021, China</span>
</div>
<!--equal contribution-->
<div class="is-size-6 publication-authors" style="margin-top: 0.4em;">
<span class="author-block">* Equal contribution</span>
<span class="author-block">† Project Lead</span>
<span class="author-block">‡ Corresponding author</span>
</div>
<!--links-->
<div class="column has-text-centered">
<div class="publication-links">
<!--arxiv link-->
<span class="link-block">
<a href="http://arxiv.org/abs/2602.19424" class="external-link button is-normal is-rounded is-dark">
<span class="icon">
<i class="ai ai-arxiv"></i>
</span>
<span>arXiv</span>
</a>
</span>
<!-- Code Link (Coming Soon) -->
<span class="link-block">
<a class="external-link button is-normal is-rounded is-dark" style="pointer-events: none; opacity: 0.6;">
<span class="icon">
<i class="fab fa-github"></i>
</span>
<span>Code (Coming Soon)</span>
</a>
</span>
<!-- Dataset Link (Coming Soon) -->
<span class="link-block">
<a class="external-link button is-normal is-rounded is-dark" style="pointer-events: none; opacity: 0.6;">
<span class="icon">
<p style="font-size:18px">🤗</p>
</span>
<span>Dataset (Coming Soon)</span>
</a>
</span>
</div>
</div>
</div>
</div>
</div>
</div>
</section>
<!----------------------------------Main Body---------------------------------->
<!-- Introduction -->
<section class="hero is-light is-small mb-0">
<div class="hero-body has-text-centered">
<h1 class="title is-1 hepato">
<span class="hepato" style="vertical-align: middle">🌟 Introduction</span>
</h1>
</div>
</section>
<section class="section">
<div class="container">
<div class="columns is-centered has-text-centered">
<div class="column is-four-fifths">
<div class="content has-text-justified">
<p>
Hepatocellular Carcinoma (HCC) relies on histopathological <b>Whole Slide Images (WSIs)</b> examination as the gold standard. However, manual analysis of these gigapixel, highly heterogeneous WSIs is labor-intensive and prone to inter-observer variability. This has catalyzed WSI-based <b>Multi-modal Large Language Models (MLLMs)</b> to enable VQA.
</p>
<p>
A key challenge in pathology MLLMs is gigapixel WSI representation. Existing methods either use <b>thumbnail-based approaches</b> that lose critical high-resolution diagnostic details, or employ <b>slide-encoder approaches</b> that generate excessively redundant tokens.
</p>
<p>
We propose <span class="hepato-brand">Hepato-LLaVA</span>, a specialized MLLM for fine-grained hepatocellular pathology analysis. It features a novel <b>Hierarchical Sparse Visual Attention (HSVA)</b> mechanism that models 2D tissue topology to aggregate diagnostic evidence while preserving context. To address multiscale data scarcity, we also present <span class="hepato-brand">HepatoPathoVQA</span>, comprising <b>33K hierarchically structured QA pairs</b> validated by pathologists. <span class="hepato-brand">Hepato-LLaVA</span> achieves state-of-the-art diagnostic accuracy, outperforming existing pathology MLLMs by an absolute <b>20%</b>.
</p>
</div>
</div>
</div>
</div>
</section>
<!--Dataset section-->
<section class="hero is-light is-small mb-0">
<div class="hero-body has-text-centered">
<h1 class="title is-1 hepato">
<!-- TODO: 替换为 HepatoPathoVQA 的图标 -->
<img src="static/images/icon.png" style="width:2em;vertical-align: middle" alt="Logo"/>
<span class="hepato-brand" style="vertical-align: middle">HepatoPathoVQA Dataset</span>
</h1>
</div>
</section>
<section class="section">
<div class="container">
<div class="columns is-centered has-text-centered">
<div class="column is-four-fifths">
<!-- Dataset Construction Pipeline -->
<h2 class="title is-3">Three-Stage Construction Pipeline</h2>
<div class="content has-text-justified">
<p>
We collected <b>200 WSIs</b> containing HCC and constructed <span class="hepato-brand">HepatoPathoVQA</span>, a multi-scale dataset featuring <b>33K QA pairs</b> for morphological analysis and diagnosis. By following pathologists' diagnostic workflows, we developed a generation pipeline using Gemini-3-flash that simulates the transition from macroscopic to microscopic clinical reasoning.
</p>
<p>
The construction pipeline consists of three stages: <b>(1) Hierarchical Sampling</b> using a Minimum Spanning Tree (MST) to identify ROIs, <b>(2) Hierarchical Clinical Inference</b> with Gemini-3-flash integrating macroscopic contexts into microscopic analysis, and <b>(3) QA Generation</b> for instruction tuning.
</p>
<p>
The dataset spans three scales—<b>WSI</b>, <b>ROI (2×)</b>, and <b>Patch (10×, 20×)</b>—validated by expert pathologists. <span class="hepato-brand">HepatoPathoVQA</span> is the first 33K multi-scale pathology dataset, bridging instruction data with real-world clinical practice.
</p>
</div>
<div class="content has-text-centered">
<img src="static/images/dataset.png" width="100%"/>
<p>
Overview of the <span class="hepato-brand">HepatoPathoVQA</span> construction pipeline:
(1) Extracts ROIs and Patches from WSIs using MST-based clustering and triangular seed-point selection.
(2) Employs Gemini-3-flash for hierarchical inference by integrating macroscopic descriptions as context for subsequent microscopic analysis.
(3) Generates multi-scale QA pairs and captions for instruction tuning and alignment.
</p>
</div>
<h2 class="title is-3">Dataset Statistics</h2>
<div class="content has-text-justified">
<p>
<span class="hepato-brand">HepatoPathoVQA</span> covers morphological assessment to final diagnosis, supporting fine-grained, multi-scale HCC pathology analysis.
</p>
</div>
<div class="box">
<div class="table-container">
<table class="table is-bordered is-striped is-narrow is-hoverable is-fullwidth has-text-centered">
<thead>
<tr>
<th>Property</th>
<th>Value</th>
</tr>
</thead>
<tbody>
<tr>
<td>Total WSIs</td>
<td><b>200</b></td>
</tr>
<tr>
<td>Total QA Pairs</td>
<td><b>33K</b></td>
</tr>
<tr>
<td>Scales</td>
<td>WSI, ROI (2×), Patch (10×, 20×)</td>
</tr>
<tr>
<td>Task Types</td>
<td>VQA (Single-choice, Multi-choice, Open-ended), Captioning</td>
</tr>
</tbody>
</table>
</div>
</div>
</div>
</div>
</div>
</section>
<!--Framework section-->
<section class="hero is-light is-small mb-0">
<div class="hero-body has-text-centered">
<h1 class="title is-1 hepato">
<!-- TODO: 替换为框架图标 -->
<img src="static/images/icon.png" style="width:2em;vertical-align: middle" alt="Logo"/>
<span class="hepato-brand" style="vertical-align: middle">Hepato-LLaVA Framework</span>
</h1>
</div>
</section>
<section class="section">
<div class="container">
<div class="columns is-centered has-text-centered">
<div class="column is-four-fifths">
<div class="content has-text-justified">
<p>
<span class="hepato-brand">Hepato-LLaVA</span> employs a modular architecture with three components: a frozen <b>Patch Encoder</b>, a novel <b>Hierarchical Sparse Visual Attention (HSVA)</b> slide encoder, and a <b>Q-Former Connector</b> compressing features into 32 learnable LLM queries.
</p>
<p>
The core <b>HSVA</b> mechanism models 2D tissue topology to explicitly aggregate local diagnostic evidence into semantic summary tokens while preserving global context. Unlike conventional MIL collapsing patches into a single token, HSVA retains spatially coherent, multi-granular representations mimicking pathologists' local-global diagnostic workflow.
</p>
<p>
The model undergoes a <b>three-stage training</b>: (1) connector pre-training on HepatoPathCaption, (2) full-model fine-tuning on HepatoPathoVQA, and (3) alignment for robust multi-scale diagnostic performance.
</p>
</div>
<div class="content has-text-centered">
<img src="static/images/model.png" width="100%"/>
<p>
Overview of the <span class="hepato-brand">Hepato-LLaVA</span> framework:
(<b>Upper</b>) Incorporates Sparse Topo-Pack Attention into the model architecture.
(<b>Lower</b>) Implements a three-stage training pipeline: MAE pre-training, MoCo pre-training, and instruction tuning.
The sparse attention mask defines three topological interactions:
(1) <b>Global Sink</b> for macro-context broadcasting,
(2) <b>Intra-Pack</b> for local dense interactions, and
(3) <b>Inter-Pack</b> for summary-level connections across packs.
</p>
</div>
</div>
</div>
</div>
</section>
<!----------------------------------Experiment Results---------------------------------->
<section class="hero is-light is-small">
<div class="hero-body has-text-centered">
<h1 class="title is-1 hepato">📊 Experiment Results</h1>
</div>
</section>
<!-- Main Comparison Table -->
<section class="section">
<div class="container">
<div class="columns is-centered">
<div class="column is-full has-text-centered content">
<h2 class="title is-3" id="leaderboard"><a>🏆 Main Results on HepatoPathoBench</a></h2>
<div class="content">
<p class="mt-3">
📈 Evaluation on <span class="hepato-brand">HepatoPathoBench</span> against general and pathology-specific MLLMs. <b>Single/Multi</b>: single/multiple choice. <b>WSI-P</b>: patch-level BLEU on WSI captioning. <b>WSI, ROI, Patch</b>: multi-scale accuracy. <b>Bold</b>: best; <u>underlined</u>: second best.
</p>
<div style="overflow-x: auto;">
<table id="table1">
<thead>
<tr class="thead-t1">
<th rowspan="3">Model</th>
<th rowspan="3">Input</th>
<th colspan="4">Morphological Analysis</th>
<th colspan="4">Diagnosis</th>
<th colspan="3">Multi-scale</th>
<th rowspan="3"><b>Avg</b></th>
</tr>
<tr class="thead-t2">
<th colspan="2">Open</th>
<th colspan="2">Close</th>
<th colspan="2">Open</th>
<th colspan="2">Close</th>
<th rowspan="2">WSI↑</th>
<th rowspan="2">ROI↑</th>
<th rowspan="2">Patch↑</th>
</tr>
<tr class="thead-t3">
<th>WSI-P↑</th>
<th>METEOR↑</th>
<th>Single↑</th>
<th>Multi↑</th>
<th>WSI-P↑</th>
<th>METEOR↑</th>
<th>Single↑</th>
<th>Multi↑</th>
</tr>
</thead>
<tbody>
<tr>
<td>Lingshu</td>
<td>Thumbnail</td>
<td>0.53</td>
<td>0.17</td>
<td>0.38</td>
<td>0.44</td>
<td><u>0.73</u></td>
<td>0.18</td>
<td>0.39</td>
<td>0.38</td>
<td>0.52</td>
<td>0.52</td>
<td>0.49</td>
<td>0.50</td>
</tr>
<tr>
<td>Huatuo-GPT</td>
<td>Thumbnail</td>
<td><u>0.74</u></td>
<td><u>0.24</u></td>
<td>0.81</td>
<td>0.45</td>
<td>0.70</td>
<td><u>0.23</u></td>
<td>0.59</td>
<td>0.32</td>
<td>0.60</td>
<td>0.65</td>
<td>0.65</td>
<td>0.65</td>
</tr>
<tr>
<td>Quilt-LLaVA</td>
<td>Thumbnail</td>
<td>0.64</td>
<td>0.22</td>
<td>0.47</td>
<td>0.32</td>
<td>0.56</td>
<td>0.15</td>
<td>0.57</td>
<td>0.37</td>
<td>0.57</td>
<td>0.60</td>
<td>0.55</td>
<td>0.57</td>
</tr>
<tr>
<td>Patho-R1</td>
<td>Thumbnail</td>
<td>0.66</td>
<td>0.19</td>
<td><u>0.87</u></td>
<td><u>0.50</u></td>
<td>0.20</td>
<td>0.05</td>
<td>0.59</td>
<td><u>0.45</u></td>
<td>0.55</td>
<td>0.55</td>
<td>0.54</td>
<td>0.55</td>
</tr>
<tr>
<td>SlideChat</td>
<td>WSI</td>
<td>0.70</td>
<td>0.17</td>
<td><u>0.87</u></td>
<td>0.47</td>
<td>0.72</td>
<td>0.14</td>
<td>0.63</td>
<td>0.39</td>
<td><u>0.66</u></td>
<td><u>0.68</u></td>
<td><u>0.66</u></td>
<td><u>0.66</u></td>
</tr>
<tr>
<td>WSI-LLaVA</td>
<td>WSI</td>
<td>0.69</td>
<td>0.20</td>
<td>0.84</td>
<td>0.46</td>
<td>0.67</td>
<td>0.16</td>
<td><u>0.65</u></td>
<td>0.36</td>
<td>0.65</td>
<td>0.67</td>
<td>0.64</td>
<td>0.65</td>
</tr>
<tr class="ours-row">
<td><b>Hepato-LLaVA 🏆</b></td>
<td><b>WSI</b></td>
<td><b>0.79</b></td>
<td><b>0.33</b></td>
<td><b>0.97</b></td>
<td><b>0.88</b></td>
<td><b>0.75</b></td>
<td><b>0.33</b></td>
<td><b>0.87</b></td>
<td><b>0.68</b></td>
<td><b>0.82</b></td>
<td><b>0.83</b></td>
<td><b>0.83</b></td>
<td><b>0.83</b></td>
</tr>
</tbody>
</table>
</div>
</div>
</div>
</div>
</div>
</section>
<!----------------------------------Case Studies---------------------------------->
<style>
.case-img-slider { position: relative; width: 100%; text-align: center; }
.case-img-slide { display: none; }
.case-img-slide.active { display: block; }
.case-img-slide img { max-width: 100%; border-radius: 8px; box-shadow: 0 2px 12px rgba(155,127,199,0.18); }
.case-nav-wrap { display: flex; align-items: center; justify-content: center; gap: 18px; margin-top: 14px; }
.case-nav-btn {
background: rgba(155,127,199,0.12);
border: 2px solid #9b7fc7;
color: #7c5cbf;
border-radius: 50%;
width: 42px; height: 42px;
font-size: 24px; line-height: 1;
cursor: pointer;
display: flex; align-items: center; justify-content: center;
transition: background 0.2s, color 0.2s;
padding: 0;
}
.case-nav-btn:hover { background: #9b7fc7; color: #fff; }
.case-nav-indicator { font-size: 15px; color: #555; min-width: 44px; text-align: center; user-select: none; }
</style>
<section class="hero is-light is-small">
<div class="hero-body has-text-centered">
<h1 class="title is-1">🔡 Case Study</h1>
</div>
</section>
<!-- WSI Level -->
<section class="section pb-3">
<div class="container is-max-desktop">
<div class="columns is-centered">
<div class="column is-full">
<h2 class="title is-4 has-text-centered mb-2">
🖼️ WSI Level
</h2>
<p class="has-text-centered mb-4" style="color:#555;">
Whole Slide Image-level open-ended diagnosis in HepatoPathoVQA Dataset.
</p>
<div class="has-text-centered">
<div class="case-img-slider" id="case-slider-wsi">
<div class="case-img-slide active">
<img src="static/images/wsi_1.png" alt="WSI Level Case Study 1"/>
</div>
<div class="case-img-slide">
<img src="static/images/wsi_2.png" alt="WSI Level Case Study 2"/>
</div>
<div class="case-nav-wrap">
<button class="case-nav-btn" onclick="caseSlide('wsi', -1)">‹</button>
<span class="case-nav-indicator" id="case-ind-wsi">1 / 2</span>
<button class="case-nav-btn" onclick="caseSlide('wsi', 1)">›</button>
</div>
</div>
</div>
</div>
</div>
</div>
</section>
<!-- ROI Level -->
<section class="section pb-3 pt-3">
<div class="container is-max-desktop">
<div class="columns is-centered">
<div class="column is-full">
<h2 class="title is-4 has-text-centered mb-2">
🔍 ROI Level
</h2>
<p class="has-text-centered mb-4" style="color:#555;">
Region-of-Interest level analysis examples in HepatoPathoVQA Dataset.
</p>
<div class="has-text-centered">
<div class="case-img-slider" id="case-slider-roi">
<div class="case-img-slide active">
<img src="static/images/roi_1.png" alt="ROI Level Case Study 1"/>
</div>
<div class="case-img-slide">
<img src="static/images/roi_2.png" alt="ROI Level Case Study 2"/>
</div>
<div class="case-nav-wrap">
<button class="case-nav-btn" onclick="caseSlide('roi', -1)">‹</button>
<span class="case-nav-indicator" id="case-ind-roi">1 / 2</span>
<button class="case-nav-btn" onclick="caseSlide('roi', 1)">›</button>
</div>
</div>
</div>
</div>
</div>
</div>
</section>
<!-- Patch Level -->
<section class="section pt-3">
<div class="container is-max-desktop">
<div class="columns is-centered">
<div class="column is-full">
<h2 class="title is-4 has-text-centered mb-2">
🧩 Patch Level
</h2>
<p class="has-text-centered mb-4" style="color:#555;">
Patch-level morphological analysis examples in HepatoPathoVQA Dataset.
</p>
<div class="has-text-centered">
<div class="case-img-slider" id="case-slider-patch">
<div class="case-img-slide active">
<img src="static/images/patch_1.png" alt="Patch Level Case Study 1"/>
</div>
<div class="case-img-slide">
<img src="static/images/patch_2.png" alt="Patch Level Case Study 2"/>
</div>
<div class="case-nav-wrap">
<button class="case-nav-btn" onclick="caseSlide('patch', -1)">‹</button>
<span class="case-nav-indicator" id="case-ind-patch">1 / 2</span>
<button class="case-nav-btn" onclick="caseSlide('patch', 1)">›</button>
</div>
</div>
</div>
</div>
</div>
</div>
</section>
<!----------------------------------Citation---------------------------------->
<section class="section" id="BibTeX">
<div class="container is-max-desktop content">
<h2 class="title">📚 BibTeX Citation</h2>
<pre><code>
@article{hepatollava2026,
title={Hepato-LLaVA: An Expert MLLM with Sparse Topo-Pack Attention for Hepatocellular Pathology Analysis on Whole Slide Images},
author={Yang, Yuxuan and Yan, Zhonghao and Zhang, Yi and Yun, Bo and Diao, Muxi and Zhao, Guowei and Liang, Kongming and Li, Wenbin and Ma, Zhanyu},
year={2026}
}
</code></pre>
</div>
</section>
<!--footer-->
<footer class="footer">
<div class="columns is-centered">
<div class="column is-8">
<div class="content">
<p>
This website is adapted from <a href="https://we-math.github.io/">We-Math</a> and <a href="https://csbench.github.io/">CS-Bench</a>, licensed under a <a rel="license"
href="http://creativecommons.org/licenses/by-sa/4.0/">Creative
Commons Attribution-ShareAlike 4.0 International License</a>.
</p>
</div>
</div>
</div>
</footer>
<!----------------------------------End Body---------------------------------->
<script>
function caseSlide(level, dir) {
var container = document.getElementById('case-slider-' + level);
var slides = container.querySelectorAll('.case-img-slide');
var current = Array.prototype.findIndex.call(slides, function(s) { return s.classList.contains('active'); });
slides[current].classList.remove('active');
current = (current + dir + slides.length) % slides.length;
slides[current].classList.add('active');
document.getElementById('case-ind-' + level).textContent = (current + 1) + ' / ' + slides.length;
}
</script>
</body>
</html>