PCMMD: Plasma Cells for Multiple Myeloma Diagnosis

PCMMD (Plasma Cells for Multiple Myeloma Diagnosis) is an important resource for training and validating Artificial Intelligence (AI) approaches to automate cell identification, supporting studies on MM in developing countries, and discovering approaches regarding other diseases affecting various cell types in diverse populations. The dataset is organized into three sets of cells stained with Wright-Giemsa:

- The first set contains 3,546 cells, with 54% labeled as non-plasma and 46% as plasma cells;
- The second set includes 2,021 cells organized by patients, along with their respective diagnoses;
- The third set comprises 1,615 plasma and 1,931 non-plasma cells, individually segmented without background.

All sets were used to train Deep Neural Network models, whose details and parameters are detailed in the following repository:

https://github.com/LabIA-UFBA/MMDB

PCMMD（多发性骨髓瘤诊断浆细胞数据集，Plasma Cells for Multiple Myeloma Diagnosis）是用于训练与验证人工智能（Artificial Intelligence，AI）自动化细胞识别算法的重要资源，可为发展中国家的多发性骨髓瘤（Multiple Myeloma，MM）相关研究提供支撑，并助力探索影响不同人群多种细胞类型的其他疾病的诊疗策略。该数据集包含三组经瑞氏-吉姆萨（Wright-Giemsa）染色的细胞样本：
- 第一组包含3546个细胞，其中54%被标记为非浆细胞，46%为浆细胞；
- 第二组包含2021个细胞，按患者分组并附带各自的诊断信息；
- 第三组包含1615个浆细胞与1931个非浆细胞，均为经过单独分割且无背景干扰的细胞。
上述所有数据集均用于训练深度神经网络（Deep Neural Network，DNN）模型，这些模型的详细信息与参数可在以下仓库中查阅：
https://github.com/LabIA-UFBA/MMDB

搜集汇总

背景与挑战

背景概述

PCMMD数据集是一个用于多发性骨髓瘤诊断的血浆细胞图像资源，包含三个子集：3,546个标记细胞、2,021个按患者组织的细胞以及1,615个血浆细胞和1,931个非血浆细胞的单独分割图像。该数据集旨在支持AI方法在细胞识别和疾病诊断中的应用，特别关注发展中国家的研究需求。