基于深度学习的皮肤器官芯片表皮模型图像的分类研究
徐成晓

东南大学生物科学与医学工程学院，江苏南京，210000；

摘要：本研究旨在利用深度学习技术实现皮肤器官芯片表皮模型图像的自动化分类，以解决人工标注成本高及小样本数据导致的过拟合问题。通过构建包含600张正常与异常样本的图像数据集，并采用多维度数据增强策略扩展至1200张样本缓解数据不足。选用改进型ResNeSt网络模型，结合Split-Attention机制与分组卷积优化特征表达，同时采用迁移学习策略和AdamW优化器进行模型训练。验证了其在皮肤器官芯片表皮模型图像分类中的优越性并为皮肤生物学研究与药物开发提供了高效的自动化分类工具。

关键词：深度学习；皮肤器官芯片；图像分类

参考文献

[1]Hong Z X,Zhu S T,Li H,et al.Bioengineered skin organoids:from development to applications[J].Military Medical Research,2023,10(1):40.

[2]Lee J,Van Der Valk W H,Serdy S A,Deakin C,Kim J,Le A P,Koehler K R.Generation and characterization of hair-bearing skin organoids from human pluripotent stem cells[J].Nature protocols,2022,17(5):1266-1305.

[3]Ebner-Peking P,Krisch L,Wolf M,et al.Self-assembly of differentiated progenitor cells facilitates spheroid human skin organoid formation and planar skin regeneration[J].Theranostics,2021,11(17):8430.

[4]Ramovs V,Janssen H,Fuentes I,et al.Characterization of the epidermal-dermal junction in hiPSC-derived skin organoids[J].Stem Cell Reports,2022,17(6):1279-1288.

[5]Sutterby E,Thurgood P,Baratchi S,et al.Microfluidic skin-on-a-chip models:Toward biomimetic artificial skin[J].Small,2020,16(39):2002515.

[6]Ismayilzada N,Tarar C,Dabbagh S R,et al.Skin-on-a-chip technologies towards clinical translation and commercialization[J].Biofabrication,2024,16(4):042001.

[7]Michielon E, Boninsegna M, Waaijman T, et al. Environmentally Controlled Microfluidic System Enabling Immune Cell Flow and Activation in an Endothelialised Skin-On-Chip[J]. Advanced Healthcare Materials, 2024, 13(29): 2400750.

[8]陈健,王珍妮,夏开建,等.基于深度学习的结直肠息肉内镜图像分割和分类方法比较[J].上海交通大学学报(医学版),2024,44(6):762.

[9]Alharith R, Ibrahim A O, Wahid N, et al. Comparative Analysis of ResNet Models for Skin Cancer Diagnosis: Performance Evaluation and Insights[C]//International Conference on Soft Computing and Data Mining. Cham: Springer Nature Switzerland, 2024: 370-378.

[10]Indumathi C P, Santhoshsivan V, Selvakumar R. ResNet and ResNeSt-Based Deep-Learning[C]//Digital Geography: Proceedings of the International Conference on Internet and Modern Society (IMS 2023). Springer Nature, 2024: 215.

[11]夏平,张光一,雷帮军,等.多尺度 ResNeSt-50聚合网络与置信度传播的息肉图像分割[J].光学精密工程,2023,31(18):2765-2780.

[12]Zhang H, Wu C, Zhang Z, et al. Resnest: Split-attention networks[C]//Proceedings of the IEEE/CVF conference on computer vision and pattern recognition. 2022: 2736-2746.

[13]Jia Y, Wan G, Liu L, et al. Split-attention networks with self-calibrated convolution for moon impact crater detection from multi-source data[J].Remote Sensing, 2021, 13(16): 3193.