Anti-cancer immunotherapy dramatically changes the clinical management of many types of tumours towards less harmful and more personalized treatment plans than conventional chemotherapy or radiation. Precise analysis of the spatial distribution of immune cells in the tumourous tissue is necessary to select patients that would best respond to the treatment. Here, we introduce a deep learning-based workflow for cell nuclei segmentation and subsequent immune cell identification in routine diagnostic images. We applied our workflow on a set of hematoxylin and eosin (H&E) stained breast cancer and colorectal cancer tissue images to detect tumour-infiltrating lymphocytes. Firstly, to segment all nuclei in the tissue, we applied the multiple-image input layer architecture (Micro-Net, Dice coefficient (DC) $0.79\pm 0.02$). We supplemented the Micro-Net with an introduced texture block to increase segmentation accuracy (DC = $0.80\pm 0.02$). We preserved the shallow architecture of the segmentation network with only 280 K trainable parameters (e.g. U-net with ∼1900 K parameters, DC = $0.78\pm 0.03$). Subsequently, we added an active contour layer to the ground truth images to further increase the performance (DC = $0.81\pm 0.02$). Secondly, to discriminate lymphocytes from the set of all segmented nuclei, we explored multilayer perceptron and achieved a 0.70 classification f-score. Remarkably, the binary classification of segmented nuclei was significantly improved (f-score = 0.80) by colour normalization. To inspect model generalization, we have evaluated trained models on a public dataset that was not put to use during training. We conclude that the proposed workflow achieved promising results and, with little effort, can be employed in multi-class nuclei segmentation and identification tasks. PDF  XML

中文翻译：

---

完整切片组织学图像中细胞核分割和淋巴细胞鉴定的深度学习模型

与传统的化学疗法或放射疗法相比，抗癌免疫疗法极大地改变了许多类型肿瘤的临床管理，朝着危害更小，更具个性化的治疗计划发展。精确分析肿瘤组织中免疫细胞的空间分布对于选择对治疗反应最佳的患者很有必要。在这里，我们介绍了基于深度学习的工作流，用于常规诊断图像中的细胞核分割和后续免疫细胞识别。我们将工作流程应用于苏木精和曙红（H＆E）染色的乳腺癌和结直肠癌组织图像，以检测浸润肿瘤的淋巴细胞。首先，为了分割组织中的所有核，我们应用了多图像输入层体系结构（Micro-Net，骰子系数（DC）$ 0.79 \ pm 0.02 $）。我们为Micro-Net补充了引入的纹理块，以提高分割精度（DC = $ 0.80 \ pm 0.02 $）。我们保留了只有280 K可训练参数的分段网络的浅层架构（例如，具有约1900 K参数的U-net，DC = $ 0.78 \ pm 0.03 $）。随后，我们向地面真实图像添加了一个活动轮廓层，以进一步提高性能（DC = $ 0.81 \ pm 0.02 $）。其次，为了区分所有分段核中的淋巴细胞，我们探索了多层感知器并获得了0.70分类的f评分。值得注意的是，通过颜色归一化可以显着改善分段核的二进制分类（f分数= 0.80）。为了检查模型的概括性，我们已经在训练期间未使用的公共数据集上评估了训练后的模型。我们得出的结论是，所提出的工作流程取得了可喜的成果，并且只需花费很少的精力，即可用于多类核的分割和识别任务。PDF XML