OSCR

CAHA-Net: A novel MR image classification model based on DenseNet incorporating coordinate attention and hybrid augmentation.

Code ↔ Paper

3 matches between paragraphs of the paper and lines of its authors' code, computed by the harvester (lexical-v1). Click a colored paragraph or line to see its counterpart.

The 3 matches
  1. [1] § Model architecture › Algorithm design ↔ neurotumornet_classifier.py, lines 25–92 · score 0.55 · BatchNorm, max pooling, dense, modules, classification, tumor
  2. [2] § Experimental results › Experimental environment and parameter settings ↔ train_swin_tiny.py, lines 336–394 · score 0.50 · weight decay, Focal Loss, Swin, Adam, brain tumor, optimization
  3. [3] § Experimental results › Experimental environment and parameter settings ↔ train_swin_tiny_partial_ft.py, lines 147–200 · score 0.50 · weight decay, Focal Loss, Swin, Adam, brain tumor, optimization

Paper

Loaded from Europe PMC by your browser, not stored by OSCR: doi.org · Europe PMC

The paper is loaded when this pane is shown.

The authors' code

Python · 470 lines · 16 KB · GPL-3.0 · 1 match

  1. """
  2. NeuroTumorNet - Brain Tumor Classification using Deep Convolutional Neural Networks
  3. ===================================================================================
  4. 基于 https://github.com/h9zdev/NeuroTumorNet 的 PyTorch 实现
  6. 原始论文方法:
  7. 1. Baseline CNN: 3层卷积块 + BatchNorm + L2正则化 + Dropout
  8. 2. VGG16 Transfer Learning: 预训练VGG16 + 自定义分类头
  9. 3. 带正则化的增强CNN: BatchNorm + Dropout + L2正则化
  11. 本模块用 PyTorch 复现原始仓库的核心架构，保留其设计理念：
  12. - 简单的3层CNN作为基线
  13. - BatchNormalization稳定训练
  14. - Dropout防止过拟合
  15. - L2正则化 (通过weight_decay实现)
  17. 作者: 基于 h9zdev/NeuroTumorNet (CC BY-NC 4.0)
  18. """
  20. import torch
  21. import torch.nn as nn
  22. import torch.nn.functional as F
  25. # ============================================================================
  26. # 1. NeuroTumorNet Baseline CNN
  27. # ============================================================================
  28. class NeuroTumorNetBaseline(nn.Module):
  29. """
  30. NeuroTumorNet Baseline CNN - 原始论文的基线模型
  32. 架构:
  33. - 3个卷积块: Conv2D -> BatchNorm -> ReLU -> MaxPool
  34. - 通道数: 32 -> 64 -> 128
  35. - 全连接层: Flatten -> Dense(128) -> Dropout -> Output
  37. 基于论文 Section 3.3: Model Architecture
  38. """
  39. def __init__(self, num_classes=4, dropout_rate=0.5):
  40. super(NeuroTumorNetBaseline, self).__init__()
  42. # 卷积块 1: 32 filters
  43. self.conv1 = nn.Conv2d(3, 32, kernel_size=3, padding=1)
  44. self.bn1 = nn.BatchNorm2d(32)
  45. self.pool1 = nn.MaxPool2d(2, 2)
  47. # 卷积块 2: 64 filters
  48. self.conv2 = nn.Conv2d(32, 64, kernel_size=3, padding=1)
  49. self.bn2 = nn.BatchNorm2d(64)
  50. self.pool2 = nn.MaxPool2d(2, 2)
  52. # 卷积块 3: 128 filters
  53. self.conv3 = nn.Conv2d(64, 128, kernel_size=3, padding=1)
  54. self.bn3 = nn.BatchNorm2d(128)
  55. self.pool3 = nn.MaxPool2d(2, 2)
  57. # 分类头
  58. # 输入 224x224 -> 经过3次pool后 -> 28x28x128
  59. self.flatten = nn.Flatten()
  60. self.fc1 = nn.Linear(128 * 28 * 28, 128)
  61. self.dropout = nn.Dropout(dropout_rate)
  62. self.fc2 = nn.Linear(128, num_classes)
  64. self._initialize_weights()
  66. def _initialize_weights(self):
  67. for m in self.modules():
  68. if isinstance(m, nn.Conv2d):
  69. nn.init.kaiming_normal_(m.weight, mode='fan_out', nonlinearity='relu')
  70. if m.bias is not None:
  71. nn.init.constant_(m.bias, 0)
  72. elif isinstance(m, nn.BatchNorm2d):
  73. nn.init.constant_(m.weight, 1)
  74. nn.init.constant_(m.bias, 0)
  75. elif isinstance(m, nn.Linear):
  76. nn.init.normal_(m.weight, 0, 0.01)
  77. nn.init.constant_(m.bias, 0)
  79. def forward(self, x):
  80. # 卷积块 1
  81. x = self.pool1(F.relu(self.bn1(self.conv1(x))))
  82. # 卷积块 2
  83. x = self.pool2(F.relu(self.bn2(self.conv2(x))))
  84. # 卷积块 3
  85. x = self.pool3(F.relu(self.bn3(self.conv3(x))))
  87. # 分类
  88. x = self.flatten(x)
  89. x = F.relu(self.fc1(x))
  90. x = self.dropout(x)
  91. x = self.fc2(x)
  92. return x
  95. # ============================================================================
  96. # 2. NeuroTumorNet Enhanced CNN (with Regularization)
  97. # ============================================================================
  98. class NeuroTumorNetEnhanced(nn.Module):
  99. """
  100. NeuroTumorNet Enhanced CNN - 带增强正则化的模型
  102. 基于论文 Section 3.6: Addressing Overfitting
  103. - 每个卷积层后添加BatchNorm
  104. - 更高的Dropout率 (可配置0.3-0.7)
  105. - L2正则化通过optimizer的weight_decay实现
  107. 架构增强:
  108. - 更深的全连接层
  109. - 多级Dropout
  110. """
  111. def __init__(self, num_classes=4, dropout_rate1=0.5, dropout_rate2=0.3):
  112. super(NeuroTumorNetEnhanced, self).__init__()
  114. # 卷积块 1
  115. self.conv1 = nn.Conv2d(3, 32, kernel_size=3, padding=1)
  116. self.bn1 = nn.BatchNorm2d(32)
  117. self.pool1 = nn.MaxPool2d(2, 2)
  119. # 卷积块 2
  120. self.conv2 = nn.Conv2d(32, 64, kernel_size=3, padding=1)
  121. self.bn2 = nn.BatchNorm2d(64)
  122. self.pool2 = nn.MaxPool2d(2, 2)
  124. # 卷积块 3
  125. self.conv3 = nn.Conv2d(64, 128, kernel_size=3, padding=1)
  126. self.bn3 = nn.BatchNorm2d(128)
  127. self.pool3 = nn.MaxPool2d(2, 2)
  129. # 增强的分类头
  130. self.flatten = nn.Flatten()
  131. self.fc1 = nn.Linear(128 * 28 * 28, 256)
  132. self.bn_fc1 = nn.BatchNorm1d(256)
  133. self.dropout1 = nn.Dropout(dropout_rate1)
  134. self.fc2 = nn.Linear(256, 128)
  135. self.dropout2 = nn.Dropout(dropout_rate2)
  136. self.fc3 = nn.Linear(128, num_classes)
  138. self._initialize_weights()
  140. def _initialize_weights(self):
  141. for m in self.modules():
  142. if isinstance(m, nn.Conv2d):
  143. nn.init.kaiming_normal_(m.weight, mode='fan_out', nonlinearity='relu')
  144. elif isinstance(m, (nn.BatchNorm2d, nn.BatchNorm1d)):
  145. nn.init.constant_(m.weight, 1)
  146. nn.init.constant_(m.bias, 0)
  147. elif isinstance(m, nn.Linear):
  148. nn.init.normal_(m.weight, 0, 0.01)
  149. nn.init.constant_(m.bias, 0)
  151. def forward(self, x):
  152. # 卷积块
  153. x = self.pool1(F.relu(self.bn1(self.conv1(x))))
  154. x = self.pool2(F.relu(self.bn2(self.conv2(x))))
  155. x = self.pool3(F.relu(self.bn3(self.conv3(x))))
  157. # 增强分类头
  158. x = self.flatten(x)
  159. x = F.relu(self.bn_fc1(self.fc1(x)))
  160. x = self.dropout1(x)
  161. x = F.relu(self.fc2(x))
  162. x = self.dropout2(x)
  163. x = self.fc3(x)
  164. return x
  167. # ============================================================================
  168. # 3. NeuroTumorNet Deep (5-layer CNN)
  169. # ============================================================================
  170. class NeuroTumorNetDeep(nn.Module):
  171. """
  172. NeuroTumorNet Deep - 更深的CNN变体
  174. 5个卷积块:
  175. - 通道数: 32 -> 64 -> 128 -> 256 -> 512
  176. - 使用Global Average Pooling代替Flatten减少参数
  177. """
  178. def __init__(self, num_classes=4, dropout_rate=0.5):
  179. super(NeuroTumorNetDeep, self).__init__()
  181. # 卷积块序列
  182. self.features = nn.Sequential(
  183. # Block 1
  184. nn.Conv2d(3, 32, kernel_size=3, padding=1),
  185. nn.BatchNorm2d(32),
  186. nn.ReLU(inplace=True),
  187. nn.MaxPool2d(2, 2),
  189. # Block 2
  190. nn.Conv2d(32, 64, kernel_size=3, padding=1),
  191. nn.BatchNorm2d(64),
  192. nn.ReLU(inplace=True),
  193. nn.MaxPool2d(2, 2),
  195. # Block 3
  196. nn.Conv2d(64, 128, kernel_size=3, padding=1),
  197. nn.BatchNorm2d(128),
  198. nn.ReLU(inplace=True),
  199. nn.MaxPool2d(2, 2),
  201. # Block 4
  202. nn.Conv2d(128, 256, kernel_size=3, padding=1),
  203. nn.BatchNorm2d(256),
  204. nn.ReLU(inplace=True),
  205. nn.MaxPool2d(2, 2),
  207. # Block 5
  208. nn.Conv2d(256, 512, kernel_size=3, padding=1),
  209. nn.BatchNorm2d(512),
  210. nn.ReLU(inplace=True),
  211. nn.MaxPool2d(2, 2),
  212. )
  214. # Global Average Pooling
  215. self.global_pool = nn.AdaptiveAvgPool2d(1)
  217. # 分类头
  218. self.classifier = nn.Sequential(
  219. nn.Flatten(),
  220. nn.Linear(512, 256),
  221. nn.ReLU(inplace=True),
  222. nn.Dropout(dropout_rate),
  223. nn.Linear(256, num_classes)
  224. )
  226. self._initialize_weights()
  228. def _initialize_weights(self):
  229. for m in self.modules():
  230. if isinstance(m, nn.Conv2d):
  231. nn.init.kaiming_normal_(m.weight, mode='fan_out', nonlinearity='relu')
  232. elif isinstance(m, nn.BatchNorm2d):
  233. nn.init.constant_(m.weight, 1)
  234. nn.init.constant_(m.bias, 0)
  235. elif isinstance(m, nn.Linear):
  236. nn.init.normal_(m.weight, 0, 0.01)
  237. nn.init.constant_(m.bias, 0)
  239. def forward(self, x):
  240. x = self.features(x)
  241. x = self.global_pool(x)
  242. x = self.classifier(x)
  243. return x
  246. # ============================================================================
  247. # 4. NeuroTumorNet VGG-style (类似VGG的架构)
  248. # ============================================================================
  249. class NeuroTumorNetVGGStyle(nn.Module):
  250. """
  251. NeuroTumorNet VGG-style - 模仿VGG的双卷积块设计
  253. 参考原始仓库使用VGG16进行迁移学习的思路，
  254. 但这里是从头训练一个VGG风格的小型网络
  256. 每个stage有2个连续的3x3卷积
  257. """
  258. def __init__(self, num_classes=4, dropout_rate=0.5):
  259. super(NeuroTumorNetVGGStyle, self).__init__()
  261. self.features = nn.Sequential(
  262. # Stage 1: 2 x Conv(64)
  263. nn.Conv2d(3, 64, kernel_size=3, padding=1),
  264. nn.BatchNorm2d(64),
  265. nn.ReLU(inplace=True),
  266. nn.Conv2d(64, 64, kernel_size=3, padding=1),
  267. nn.BatchNorm2d(64),
  268. nn.ReLU(inplace=True),
  269. nn.MaxPool2d(2, 2),
  271. # Stage 2: 2 x Conv(128)
  272. nn.Conv2d(64, 128, kernel_size=3, padding=1),
  273. nn.BatchNorm2d(128),
  274. nn.ReLU(inplace=True),
  275. nn.Conv2d(128, 128, kernel_size=3, padding=1),
  276. nn.BatchNorm2d(128),
  277. nn.ReLU(inplace=True),
  278. nn.MaxPool2d(2, 2),
  280. # Stage 3: 3 x Conv(256)
  281. nn.Conv2d(128, 256, kernel_size=3, padding=1),
  282. nn.BatchNorm2d(256),
  283. nn.ReLU(inplace=True),
  284. nn.Conv2d(256, 256, kernel_size=3, padding=1),
  285. nn.BatchNorm2d(256),
  286. nn.ReLU(inplace=True),
  287. nn.Conv2d(256, 256, kernel_size=3, padding=1),
  288. nn.BatchNorm2d(256),
  289. nn.ReLU(inplace=True),
  290. nn.MaxPool2d(2, 2),
  292. # Stage 4: 3 x Conv(512)
  293. nn.Conv2d(256, 512, kernel_size=3, padding=1),
  294. nn.BatchNorm2d(512),
  295. nn.ReLU(inplace=True),
  296. nn.Conv2d(512, 512, kernel_size=3, padding=1),
  297. nn.BatchNorm2d(512),
  298. nn.ReLU(inplace=True),
  299. nn.Conv2d(512, 512, kernel_size=3, padding=1),
  300. nn.BatchNorm2d(512),
  301. nn.ReLU(inplace=True),
  302. nn.MaxPool2d(2, 2),
  303. )
  305. self.avgpool = nn.AdaptiveAvgPool2d((7, 7))
  307. self.classifier = nn.Sequential(
  308. nn.Linear(512 * 7 * 7, 512),
  309. nn.ReLU(inplace=True),
  310. nn.Dropout(dropout_rate),
  311. nn.Linear(512, 256),
  312. nn.ReLU(inplace=True),
  313. nn.Dropout(dropout_rate),
  314. nn.Linear(256, num_classes)
  315. )
  317. self._initialize_weights()
  319. def _initialize_weights(self):
  320. for m in self.modules():
  321. if isinstance(m, nn.Conv2d):
  322. nn.init.kaiming_normal_(m.weight, mode='fan_out', nonlinearity='relu')
  323. elif isinstance(m, nn.BatchNorm2d):
  324. nn.init.constant_(m.weight, 1)
  325. nn.init.constant_(m.bias, 0)
  326. elif isinstance(m, nn.Linear):
  327. nn.init.normal_(m.weight, 0, 0.01)
  328. nn.init.constant_(m.bias, 0)
  330. def forward(self, x):
  331. x = self.features(x)
  332. x = self.avgpool(x)
  333. x = torch.flatten(x, 1)
  334. x = self.classifier(x)
  335. return x
  338. # ============================================================================
  339. # 5. NeuroTumorNet Tiny (轻量级版本)
  340. # ============================================================================
  341. class NeuroTumorNetTiny(nn.Module):
  342. """
  343. NeuroTumorNet Tiny - 轻量级版本用于快速实验
  345. 更少的通道数: 16 -> 32 -> 64
  346. 使用Global Average Pooling大幅减少参数
  347. """
  348. def __init__(self, num_classes=4, dropout_rate=0.3):
  349. super(NeuroTumorNetTiny, self).__init__()
  351. self.features = nn.Sequential(
  352. # Block 1
  353. nn.Conv2d(3, 16, kernel_size=3, padding=1),
  354. nn.BatchNorm2d(16),
  355. nn.ReLU(inplace=True),
  356. nn.MaxPool2d(2, 2),
  358. # Block 2
  359. nn.Conv2d(16, 32, kernel_size=3, padding=1),
  360. nn.BatchNorm2d(32),
  361. nn.ReLU(inplace=True),
  362. nn.MaxPool2d(2, 2),
  364. # Block 3
  365. nn.Conv2d(32, 64, kernel_size=3, padding=1),
  366. nn.BatchNorm2d(64),
  367. nn.ReLU(inplace=True),
  368. nn.MaxPool2d(2, 2),
  370. # Block 4
  371. nn.Conv2d(64, 128, kernel_size=3, padding=1),
  372. nn.BatchNorm2d(128),
  373. nn.ReLU(inplace=True),
  374. nn.AdaptiveAvgPool2d(1),
  375. )
  377. self.classifier = nn.Sequential(
  378. nn.Flatten(),
  379. nn.Linear(128, 64),
  380. nn.ReLU(inplace=True),
  381. nn.Dropout(dropout_rate),
  382. nn.Linear(64, num_classes)
  383. )
  385. self._initialize_weights()
  387. def _initialize_weights(self):
  388. for m in self.modules():
  389. if isinstance(m, nn.Conv2d):
  390. nn.init.kaiming_normal_(m.weight, mode='fan_out', nonlinearity='relu')
  391. elif isinstance(m, nn.BatchNorm2d):
  392. nn.init.constant_(m.weight, 1)
  393. nn.init.constant_(m.bias, 0)
  394. elif isinstance(m, nn.Linear):
  395. nn.init.normal_(m.weight, 0, 0.01)
  396. nn.init.constant_(m.bias, 0)
  398. def forward(self, x):
  399. x = self.features(x)
  400. x = self.classifier(x)
  401. return x
  404. # ============================================================================
  405. # 模型工厂函数
  406. # ============================================================================
  407. def create_neurotumornet(variant='baseline', num_classes=4, **kwargs):
  408. """
  409. 创建 NeuroTumorNet 模型
  411. Args:
  412. variant: 模型变体
  413. - 'baseline': 原始3层CNN
  414. - 'enhanced': 带增强正则化的CNN
  415. - 'deep': 5层深度CNN
  416. - 'vgg_style': VGG风格的CNN
  417. - 'tiny': 轻量级CNN
  418. num_classes: 输出类别数
  419. **kwargs: 额外参数 (如 dropout_rate)
  421. Returns:
  422. PyTorch模型
  423. """
  424. models = {
  425. 'baseline': NeuroTumorNetBaseline,
  426. 'enhanced': NeuroTumorNetEnhanced,
  427. 'deep': NeuroTumorNetDeep,
  428. 'vgg_style': NeuroTumorNetVGGStyle,
  429. 'tiny': NeuroTumorNetTiny,
  430. }
  432. if variant not in models:
  433. raise ValueError(f"Unknown variant: {variant}. Choose from {list(models.keys())}")
  435. return models[variant](num_classes=num_classes, **kwargs)
  438. # ============================================================================
  439. # 测试代码
  440. # ============================================================================
  441. if __name__ == '__main__':
  442. print("=" * 60)
  443. print("测试 NeuroTumorNet Classifier")
  444. print("=" * 60)
  446. device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
  447. dummy_input = torch.randn(2, 3, 224, 224).to(device)
  449. models_to_test = [
  450. ('NeuroTumorNetBaseline', 'baseline'),
  451. ('NeuroTumorNetEnhanced', 'enhanced'),
  452. ('NeuroTumorNetDeep', 'deep'),
  453. ('NeuroTumorNetVGGStyle', 'vgg_style'),
  454. ('NeuroTumorNetTiny', 'tiny'),
  455. ]
  457. for name, variant in models_to_test:
  458. print(f"\n--- {name} ---")
  459. model = create_neurotumornet(variant=variant, num_classes=4).to(device)
  460. output = model(dummy_input)
  462. params = sum(p.numel() for p in model.parameters()) / 1e6
  463. print(f"输入: {dummy_input.shape}")
  464. print(f"输出: {output.shape}")
  465. print(f"参数量: {params:.2f}M", end=" ")
  466. print("✅ 测试通过!")
  468. print("\n" + "=" * 60)
  469. print("所有测试完成!")
  470. print("=" * 60)

neurotumornet_classifier.py at commit acbe901, under GPL-3.0 · at the source

Overview

Authors: Hui Qi1, Ruizhe Guo2,3, Jiajun He2, Xiyi Wang4, Yaling Han2, Xuchu Jiang2, Chibiao Liu1,5,6
  1. School of Information Engineering, Sanming University,Sanming, 365004 China
  2. School of Statistics and Mathematics, Zhongnan University of Economics and Law,Wuhan, 430073 China
  3. Department of Statistics, University of Warwick,Coventry, CV4 7AL UK
  4. School of Mathematics, Sun Yat-sen University,Guangzhou, 510275 China
  5. Fujian Key Lab of Agriculture IOT Application, Sanming University,Sanming, 365004 China
  6. Fujian Provincial Universities Key Laboratory of Industrial Big Data Analysis and Application, Sanming University,Sanming, 365004 China
Institutions: Sanming University (China); Zhongnan University of Economics and Law (China); University of Warwick (United Kingdom); Sun Yat-sen University (China)
Journal: Scientific reports, volume 16, issue 1, article 17159
Dates: received 7 January 2026; accepted 22 May 2026; published online 28 May 2026
Type: Research article · Language: English
License: CC BY-NC-ND
Identifiers: DOI 10.1038/s41598-026-55136-1 · PMID 42209768 · PMCID PMC13234125 · OpenAlex W7162650079
Open access: gold, a free copy (OpenAlex)
Status: code verified
Categories: structural MRI / diffusion (modality), human (organism), other condition (population)
Methods: Machine learning, Statistics
Keywords: Brain tumor, Image classification, Data augmentation, Coordinate attention mechanism, DenseNet, Cancer, Computational biology and bioinformatics, Engineering, Mathematics and computing
MeSH: Brain Neoplasms*, Image Interpretation, Computer-Assisted*, Magnetic Resonance Imaging*, Algorithms, Convolutional Neural Networks, Diagnosis, Computer-Assisted, Humans, Image Processing, Computer-Assisted (* major topic)
Topic: Brain Tumor Detection and Classification (Neurology, Neuroscience), according to OpenAlex
Funding: Fujian Provincial Natural Science Foundation of China (2024J01903); Key Project of the Education Department of Fujian Province (JZ230054)
Citations: not cited yet (Europe PMC); 32 references in the paper

Abstract

The abstract is not reproduced here: the paper's license (CC BY-NC-ND) does not allow it. Read it in the paper, at the publisher or on Europe PMC.

Repository

Its files are read in the Code ↔ Paper reader above, with 3 matches between paragraphs and lines of code.

Jiajun-H/CAHA-Net

License: GPL-3.0
State: the link answers, verified on 28 September 2026
Evidence: files inventoried
Commit: acbe9011a2f9b1677492b860d04a4dab7a0d64c0, 1 May 2026
Languages: Python (51)
Size: 13,765 files, 51 scripts
Software Heritage: not archived
Found in: “Data availability”
Holds: README, license file
Not found: CITATION.cff, environment file, tests, continuous integration, documentation
Tools: PyTorch (41 files), NumPy (38 files), Pillow (32 files), scikit-learn (31 files), Matplotlib (4 files), OpenCV (2 files), SciPy (2 files), NiBabel (1 file), seaborn (1 file)
Availability: 1 check, the latest on 28 September 2026: the link answers
  • 28 September 2026: the link answers
53 files
At the source: github.com/Jiajun-H/CAHA-Net

The paper's code and data availability statement is in the Data section.

Tracing map

Proposed by the machine: these links were found in the paper and verified at the source, without human review. The map will receive a Zenodo DOI once one of the paper's authors has validated it with their ORCID.

What the map holds:

  • 1 repository of the authors' code, each at its verified commit, with its license and how the link was found in the paper;
  • 51 scripts, each with its path and the digest of its content;
  • 3 matches between paragraphs of the paper and lines of the code (method lexical-v1);
  • neither the text of the paper nor the code itself.

Its JSON (tracing-map.json) is deposited on Zenodo with its DOI once the map is validated.

Data

No dataset and no data link were found in the paper.

Code and data availability statement

The paper has a code and data availability statement. Its license (CC BY-NC-ND) does not allow reproducing it here; in short, from what the harvester recognized in it:

Read it in the paper: doi.org/10.1038/s41598-026-55136-1.

Versions

The history of this record: each version stored by the harvester or made by a correction of its authors or of the maintainers of its code, and what changed in its facts. The texts of the paper (its abstract, its availability statements) are not part of it; versions that changed only those are not listed.

Version 1, 28 September 2026: the first record

Recorded: type, language, journal, volume, issue, pages, dates, 7 authors, 9 keywords, 8 MeSH terms, 2 funders, 16 references.

Cite

This paper

Qi, H., Guo, R., He, J., Wang, X., Han, Y., Jiang, X., & Liu, C. (2026). CAHA-Net: A novel MR image classification model based on DenseNet incorporating coordinate attention and hybrid augmentation. Scientific reports, 16(1), 17159. https://doi.org/10.1038/s41598-026-55136-1

BibTeX

@article{qi2026caha,
author = {Qi, Hui and Guo, Ruizhe and He, Jiajun and Wang, Xiyi and Han, Yaling and Jiang, Xuchu and Liu, Chibiao},
title = {{CAHA-Net: A novel MR image classification model based on DenseNet incorporating coordinate attention and hybrid augmentation}},
journal = {Scientific reports},
year = {2026},
month = may,
volume = {16},
number = {1},
pages = {17159},
publisher = {Nature Publishing Group},
issn = {2045-2322},
doi = {10.1038/s41598-026-55136-1},
url = {https://doi.org/10.1038/s41598-026-55136-1},
pmid = {42209768},
pmcid = {PMC13234125}
}

RIS

TY - JOUR
AU - Qi, Hui
AU - Guo, Ruizhe
AU - He, Jiajun
AU - Wang, Xiyi
AU - Han, Yaling
AU - Jiang, Xuchu
AU - Liu, Chibiao
TI - CAHA-Net: A novel MR image classification model based on DenseNet incorporating coordinate attention and hybrid augmentation
T2 - Scientific reports
J2 - Sci Rep
PY - 2026
DA - 2026/05/28
VL - 16
IS - 1
SP - 17159
SN - 2045-2322
PB - Nature Publishing Group
DO - 10.1038/s41598-026-55136-1
UR - https://doi.org/10.1038/s41598-026-55136-1
LA - en
ER -

CSL-JSON

{
"id": "10.1038/s41598-026-55136-1",
"type": "article-journal",
"title": "CAHA-Net: A novel MR image classification model based on DenseNet incorporating coordinate attention and hybrid augmentation",
"container-title": "Scientific reports",
"author": [
{
"family": "Qi",
"given": "Hui"
},
{
"family": "Guo",
"given": "Ruizhe"
},
{
"family": "He",
"given": "Jiajun"
},
{
"family": "Wang",
"given": "Xiyi"
},
{
"family": "Han",
"given": "Yaling"
},
{
"family": "Jiang",
"given": "Xuchu"
},
{
"family": "Liu",
"given": "Chibiao"
}
],
"container-title-short": "Sci Rep",
"volume": "16",
"issue": "1",
"page": "17159",
"DOI": "10.1038/s41598-026-55136-1",
"PMID": "42209768",
"PMCID": "PMC13234125",
"ISSN": "2045-2322",
"publisher": "Nature Publishing Group",
"URL": "https://doi.org/10.1038/s41598-026-55136-1",
"language": "en",
"issued": {
"date-parts": [
[
2026,
5,
28
]
]
}
}

The tracing map gets a citation of its own once an author has validated it and it has a DOI.

Similar papers

The papers with a page that share the most with this one: the tools found in their code, their categories, datasets, cited references and authors, the rarest counting most.

[1] doi:10.3389/frai.2026.1771088 [code]
Few-shot deployment of pretrained MRI transformers in brain imaging tasks.
Journal: Frontiers in artificial intelligence
In common: OpenCV, Pillow, NiBabel, 6 other tools, structural MRI / diffusion, 1 reference
[2] doi:10.21037/qims-2026-0792 [code]
An nnU-Net-based framework with adaptive feature representation for 3D brain tumor segmentation.
Journal: Quantitative imaging in medicine and surgery
In common: OpenCV, NiBabel, PyTorch, 5 other tools, structural MRI / diffusion, other condition, 2 references
[3] doi:10.1002/alz.71649 [code]
Postmortem brain MRI reveals differential associations of subcortical and limbic volumes with cortical thinning and neurodegenerative pathologies.
Journal: Alzheimer's & dementia : the journal of the Alzheimer's Association
In common: OpenCV, Pillow, NiBabel, 6 other tools, structural MRI / diffusion, other condition
[4] doi:10.1186/s12880-026-02481-2 [code]
Deep learning-based neuroanatomical profiling reveals population-specific brain changes in multiple sclerosis: a large-scale Middle Eastern study.
Journal: BMC medical imaging
In common: OpenCV, Pillow, NiBabel, 5 other tools, structural MRI / diffusion, 1 reference
[5] doi:10.1016/j.patter.2026.101538 [code]
A multi-modal foundation model for brain disease diagnosis and medical imaging.
Journal: Patterns (New York, N.Y.)
In common: OpenCV, Pillow, NiBabel, 5 other tools, 1 reference
[6] doi:10.1038/s41598-026-57519-w [code]
Automated segmentation of neurons and spinal cord structures in immunofluorescence images using SpineDL.
Journal: Scientific reports
In common: OpenCV, Pillow, NiBabel, 5 other tools, other condition, 1 reference
[7] doi:10.1016/j.phro.2026.101056 [code]
Toward uncertainty-aware manual delineation of brain tumours using eye-tracking and image-derived features.
Journal: Physics and imaging in radiation oncology
In common: OpenCV, Pillow, NiBabel, 6 other tools, other condition
[8] doi:10.1364/boe.600665 [code]
NeuroSeg-MF: robust neuron segmentation in two-photon Ca<sup>2+</sup> imaging using multi-feature fusion and detection-guided SAM.
Journal: Biomedical optics express
In common: OpenCV, Pillow, PyTorch, 4 other tools, 2 references
[9] doi:10.1162/imag.a.1309 [code]
Probing the content of semantic representations in body-selective regions.
Journal: Imaging neuroscience (Cambridge, Mass.)
In common: OpenCV, Pillow, NiBabel, 6 other tools
[10] doi:10.1167/jov.26.5.7 [code]
Representations in vision and language converge in a shared, multidimensional space of perceived similarities.
Journal: Journal of vision
In common: OpenCV, Pillow, NiBabel, 6 other tools

Contribute

The authors of this paper can claim it, correct its record and validate its tracing map, and the maintainers of its code (its owner, or a public member of its organization) correct what it says of their repository; anyone signed in can ask for its removal. Every request goes to OSCR's own machine, which answers it; your account page follows them.

Sign in with ORCID to claim this paper as one of its authors, correct its record or validate its tracing map: when the paper's metadata lists your ORCID iD, you are recognized at once. Maintainers of its code: sign in with GitHub, then claim the repository on your account page.

Request its removal

To ask OSCR to remove this record, the copies of its authors' scripts or its tracing map, use the removal request page: signed in, you say who you are, what to remove and why, then review and confirm the request. Published rules decide every request (how).

Discussion, reproductions, activity

Discussion: questions and error reports about this paper and its code, from signed-in readers and its authors. It opens with sign-in.

Reproductions: reports from readers who ran the authors' code: what they reproduced, with which environment, commit and data. It opens with sign-in.

Activity: what happens around this paper: new versions of its record, its map's validation, discussions and reproductions. It opens with sign-in.