基于交互注意力与特征融合的胸片分类算法

Chest X-ray image classification algorithm based on interactive attention and feature fusion

摘要

针对胸部X光片分类算法特征提取不充分、细节信息丢失的问题，提出一种基于交互注意力和特征融合改进ConvNeXt的胸部X光片图像分类模型。引入可变形交互注意力模块，通过双路径特征解耦与交互调制机制，实现不同属性特征信息的协同增强；同时，构建多尺度特征融合模块，用于融合不同抽象层级的判别信息，提升模型留存细节信息的能力；此外，还采用加权焦点损失函数弥补传统交叉熵函数的不足。在ChestX-Ray14数据集上进行实验，结果表明，该方法在14种疾病分类的平均AUC达到0.851，和现有的深度学习模型相比具有一定的竞争力。本文方法可提高分类精度，为胸部X光片的辅助诊断提供有力的技术支持。

Abstract

An improved ConvNeXt-based model for chest X-ray image classification that integrates interactive attention and feature fusion is proposed to address the issues of insufficient feature extraction and detailed information loss in the existing chest X-ray image classification algorithms. A deformable interactive attention module is introduced to synergistically enhance features with different attributes through dual-path feature decoupling and interactive modulation mechanisms. Meanwhile, a multi-scale feature fusion module is constructed to integrate discriminative information across different abstraction levels, thereby improving the model's ability to retain fine-grained details. Additionally, a weighted focal loss function is adopted to compensate for the limitations of traditional cross-entropy loss. Experiments on the ChestX-Ray14 dataset demonstrate that the proposed method achieves an average AUC of 0.851 for the classification of 14 diseases, exhibiting competitive performance compared with existing deep learning models. The proposed approach can improve classification accuracy, and provide robust technical support for the auxiliary diagnosis of chest X-rays.

关键词

胸部X光片 / 深度学习 / 可变形交互注意力 / 多尺度特征融合

Key words

chest X-ray image / deep learning / deformable interactive attention / multi-scale feature fusion

参考文献

[1] Gefter WB, Post BA, Hatabu H . Commonly missed findings on chest radiographs: causes and consequences[J]. Chest, 2023, 163(3): 650-661.

[2] Moifo B, Pefura—Yone EW, Nguefack—Tsague G, et al. Inter—observer variability in the detection and interpretation of chest X—ray anomalies in adults in an endemic tuberculosis area[J]. Open J Med Imaging, 2015, 5(3): 143-149.

[3] Agrawal T, Choudhary P . Segmentation and classification on chest radiography: a systematic survey[J]. Vis Comput, 2023, 39(3): 875-913.

[4] Elkholey NK, Nasr Mehana AM, Raafat Ali W . Validation of artificial intelligence in chest X—ray interpretation[J]. QJM, 2024, 117(S2): hcae175.947.

[5] Sajed S, Sanati A, Garcia JE, et al. The effectiveness of deep learning vs. traditional methods for lung disease diagnosis using chest X—ray images: a systematic review [J]. Appl Soft Comput, 2023, 147: 110817.

[6] 庞宇, 毕瑞, 王慧倩, 等 . 基于DuaLNet网络的胸部X光图像分类算法[J]. 生命科学仪器, 2022, 20(2): 44-50.

[7] Pang Y, Bi R, Wang HQ, et al. Research progress of antiphage defense systems in bacteria[J]. Life Science Instruments, 2022, 20(2): 44-50.

[8] Guan QJ, Huang YP . Multi—label chest X—ray image classification via category—wise residual attention learning [J]. Pattern Recognit Lett, 2020, 130: 259-266.

[9] Wang HY, Wang SS, Qin ZB, et al. Triple attention learning for classification of 14 thoracic diseases using chest radiography[J]. Med Image Anal, 2021, 67: 101846.

[10] Chen BZ, Li JX, Lu GM, et al. Label co—occurrence learning with graph convolutional networks for multi—label chest X—ray image classification[J]. IEEE J Biomed Health Inform, 2020, 24(8): 2292-2302.

[11] Öztürk Ş, Turalı MY, Çukur T . HydraViT: adaptive multi—branch transformer for multi—label disease classification from chest X—ray images[J]. Biomed Signal Process Control, 2025, 100(Part A): 106959.

[12] Liu Z, Mao HZ, Wu CY, et al. A ConvNet for the 2020s[C]// 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). Piscataway, NJ, USA: IEEE, 2022: 11966-11976.

[13] Wang XS, Peng YF, Lu L, et al. ChestX—Ray8: hospital—scale chest X—ray database and benchmarks on weakly—supervised classification and localization of common thorax diseases[C]// 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR). Piscataway, NJ, USA: IEEE, 2017: 3462-3471.

[14] Sharma R, Kamra A . A review on CLAHE based enhancement techniques[C]// 2023 6th International Conference on Contemporary Computing and Informatics (IC3I). Piscataway, NJ, USA: IEEE, 2023: 321-325.

[15] Pisano ED, Zong S, Hemminger BM, et al. Contrast limited adaptive histogram equalization image processing to improve the detection of simulated spiculations in dense mammograms[J]. J Digit Imaging, 1998, 11(4): 193-200.

[16] Liu Z, Lin YT, Cao Y, et al. Swin transformer: hierarchical vision transformer using shifted windows[C]// 2021 IEEE/CVF International Conference on Computer Vision (ICCV). Piscataway, NJ, USA: IEEE, 2021: 9992-10002.

[17] He KM, Zhang XY, Ren SQ, et al. Deep residual learning for image recognition[C]// 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR). Piscataway, NJ, USA: IEEE, 2016: 770-778.

[18] Hu J, Shen L, Sun G . Squeeze—and—excitation networks[C]// 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway, NJ, USA: IEEE, 2018: 7132-7141.

[19] Woo S, Park J, Lee JY, et al. CBAM: convolutional block attention module[C]// Computer Vision—ECCV 2018. Cham: Springer International Publishing, 2018: 3-19.

[20] Ma WP, Chen C, Ma MR, et al. An adaptive dual—supervised cross—deep dependency network for pixel—wise classification[J]. IEEE Trans Geosci Remote Sens, 2025, 63: 1-13.

[21] Xie LY, Li C, Wang ZR, et al. SHISRCNet: super—resolution and classification network for low—resolution breast cancer histopathology image[C]// Medical Image Computing and Computer Assisted Intervention—MICCAI 2023. Cham: Springer Nature Switzerland, 2023: 23-32.

[22] Lin TY, Goyal P, Girshick R, et al. Focal loss for dense object detection[C]// 2017 IEEE International Conference on Computer Vision (ICCV). Piscataway, NJ, USA: IEEE, 2017: 2999-3007.

[23] Wang L, Wang CL, Sun ZQ, et al. Class balanced loss for image classification[J]. IEEE Access, 2020, 8: 81142-81153.

[24] Paszke A, Gross S, Massa F, et al. Pytorch: an imperative style, high—performance deep learning library[C]// 33rd Conference on Neural Information Processing Systems (NeurIPS 2019). Red Hook, NY, USA: Curran Associates Inc., 2019: 1-12.

[25] Kingma DP, Ba J . Adam: a method for stochastic optimization[J]. arXiv preprint arXiv: 1412.6980, 2014.

[26] Chen K, Wang XQ, Zhang SW . Thorax disease classification based on pyramidal convolution shuffle attention neural network[J]. IEEE Access, 2022, 10: 85571-85581.

[27] Mao CS, Yao L, Luo Y . ImageGCN: multi—relational image graph convolutional networks for disease identification with chest X—rays[J]. IEEE Trans Med Imaging, 2022, 41(8): 1990-2003.

[28] 胡锦波, 聂为之, 宋丹, 等 . 可形变Transformer辅助的胸部X光影像疾病诊断模型[J]. 浙江大学学报(工学版), 2023, 57(10): 1923-1932.

[29] Hu JB, Nie WZ, Song D, et al. Chest X—ray imaging disease diagnosis model assisted by deformable Transformer[J]. Journal of Zhejiang University (Engineering Science), 2023, 57(10): 1923-1932.

[30] Li DF, Huo H, Jiao SP, et al. Automated thorax disease diagnosis using multi—branch residual attention network[J]. Sci Rep, 2024, 14(1): 11865.

[31] Selvaraju RR, Cogswell M, Das A, et al. Grad—CAM: visual explanations from deep networks via gradient—based localization [C]// 2017 IEEE International Conference on Computer Vision (ICCV). Piscataway, NJ, USA: IEEE, 2017: 618-626.

基金资助

甘肃省软科学专项(23JRZA484)