Processed Multiplexed Ion Beam Imaging (MIBI) of tumor microenvironments of 54 melanoma samples following immunotherapy

This dataset consists of processed results from a MIBI (Multiplexed Ion Beam Imaging) study of 54 patients with melanoma who received immunotherapy. MIBI is a highly multiplexed imaging technique that enables the simultaneous, quantitative detection of dozens of protein markers within tissue samples at sub-cellular resolution. It combines antibody-based staining with time-of-flight secondary ion mass spectrometry (ToF-SIMS) to analyze the tumor microenvironment. The 29-protein panel used in this MIBI study included markers for broad cell typing (e.g., CD45, SOX10), myeloid subsets (CD68, CD163), lymphocyte subsets (CD3, CD8), immune checkpoints (PD-1, PD-L1, LAG-3, TIM-3), and other markers (e.g., proliferation, structural markers). After imaging on field of view (FOV) for each patient sample, the arcsinh-transformed values of the panel proteins were used in conjunction with a deep learning-based object-detection model (trained on previously segmented MIBI studies) to localize and classify the cell types within each FOV. The 'scaled_mibi_data_w_cell_gating.csv' file contains these results, where each row is for one cell identified in an FOV. This table includes cells' spatial information (x and y-coordinates and estimated area), their 29 arcsinh-transformed protein intensity measurements, and boolean columns for their cell typing. Note that these cell types are hierarchical, falling under four main categories: tumor cells, fibroblasts, smooth muscle/myofibroblasts, and immune cells. In addition, the 'patient_info.csv' file contains some basic (non-identifying) patient characteristics (age tertile at diagnosis within the 54 patient sample, and Roman numeral stage at diagnosis), along with two clinical outcome variables: 'response_binary' and 'response_multi'. Specifically, a value of 1 in the 'response_binary' variable indicates whether the patient showed stable disease, a partial response, or remission after immunotherapy (0 then represents progressive disease); the 'response_multi' variable encodes their more granular outcome (CR = complete response, PR = partial response, SD = stable disease, PD = progressive disease).本数据集源自一项针对54名接受免疫治疗的黑色素瘤患者的多重离子束成像（Multiplexed Ion Beam Imaging, MIBI）研究的经处理结果。MIBI是一种高度多重复用成像技术，可在亚细胞分辨率下对组织样本内数十种蛋白质标志物实现同步定量检测；该技术将基于抗体的染色与飞行时间二次离子质谱（time-of-flight secondary ion mass spectrometry, ToF-SIMS）相结合，用于解析肿瘤微环境。本项MIBI研究使用的29重蛋白检测面板涵盖以下几类标志物：广谱细胞分型标志物（如CD45、SOX10）、髓系细胞亚群标志物（CD68、CD163）、淋巴细胞亚群标志物（CD3、CD8）、免疫检查点标志物（PD-1、PD-L1、LAG-3、TIM-3）以及其他功能标志物（如增殖相关标志物、结构标志物）。 对每名患者样本的成像视场（field of view, FOV）完成成像后，研究人员将检测面板蛋白的反正弦变换值，与基于深度学习训练的目标检测模型（基于此前已标注分割的MIBI研究数据训练）相结合，对每个成像视场内的细胞进行定位与分型。`scaled_mibi_data_w_cell_gating.csv`文件存储了上述分析结果，文件中每一行对应一个在成像视场内被识别的细胞。该表格包含细胞的空间信息（x、y坐标及估算面积）、29种蛋白的反正弦变换后荧光强度值，以及用于细胞分型的布尔型列。需说明的是，本次研究的细胞分型具有层级结构，可归为四大类别：肿瘤细胞、成纤维细胞、平滑肌/肌成纤维细胞以及免疫细胞。 此外，`patient_info.csv`文件包含部分非识别性的患者基础特征（54名患者样本确诊时的年龄三分位分组，以及确诊时的罗马数字肿瘤分期），同时包含两项临床结局变量：`response_binary`与`response_multi`。具体而言，`response_binary`变量取值为1时，表示患者在接受免疫治疗后达到疾病稳定、部分缓解或完全缓解（取值为0则对应疾病进展）；`response_multi`变量则对患者结局进行更精细化的编码：CR代表完全缓解（complete response）、PR代表部分缓解（partial response）、SD代表疾病稳定（stable disease）、PD代表疾病进展（progressive disease）。