Digital Pathology Dataset for Prostate Cancer Diagnosis

Links to code and <em>Patterns</em> paper: 1. Multi-lens Neural Machine (MLNM) Code 2. An AI-assisted Tool For Efficient Prostate Cancer Diagnosis in Low-grade and Low-volume Cases Digitized hematoxylin and eosin (H&amp;E)-stained whole-slide-images (WSIs) of 40 prostatectomy and 59 core needle biopsy specimens were collected from 99 prostate cancer patients at Tan Tock Seng Hospital, Singapore. There were 99 WSIs in total such that each specimen had one WSI. H&amp;E-stained slides were scanned at 40× magnification (specimen-level pixel size 0·25μm × 0·25μm) using Aperio AT2 Slide Scanner (Leica Biosystems). Institutional board review from the hospital were obtained for this study, and all the data were de-identified. Prostate glandular structures in core needle biopsy slides were manually annotated and classified using the ASAP annotation tool (ASAP). A senior pathologist reviewed 10% of the annotations in each slide, ensuring that some reference annotations were provided to the researcher at different regions of the core. It is to be noted that partial glands appearing at the edges of the biopsy cores were not annotated. <strong>Whole Slide Image Dataset </strong> Whole Slide Image dataset containing 99 images in SVS format with corresponding annotations in XML format are provided in WSI.zip. Available patient grading for the WSIs are provided in 'gleason_score_mapped.txt'. These XML annotations can be parsed using the code in official repository. <strong>Cropped Image Dataset </strong> Patches of size 512 × 512 pixels were cropped from the WSI (Whole Slide Image Dataset) at resolutions 5×, 10×, 20×, and 40× with an annotated gland centered at each patch. This dataset contains these cropped images. This dataset is used to train the two AI models for Gland Segmentation (99 patients) and Gland Classification (46 patients). Tables 1 and 2 illustrate both gland segmentation and gland classification datasets. We have put the two corresponding sub-datasets as two zip files as follows: gland_segmentation_dataset.zip gland_classification_dataset.zip <strong>Table 1:</strong> The number of slides and patches in training, validation, and test sets for gland segmentation task. There is one H&amp;E stained WSI for each prostatectomy or core needle biopsy specimen. <strong>#Slides</strong> Train Valid Test <strong>Total</strong> Prostatectomy 17 8 15 40 Biopsy 26 13 20 59 <strong>Total</strong> 43 21 35 99 <strong>#Patches</strong> Train Valid Test <strong>Total</strong> Prostatectomy 7795 3753 7224 18772 Biopsy 5559 4028 5981 15568 <strong>Total</strong> 13354 7781 13205 34340 <strong>Table 2:</strong> The number of slides and patches in training, validation, and test sets for gland classification task. There is one H&amp;E stained WSI for each prostatectomy or core needle biopsy specimen. The gland classification datasets are the subsets of the gland segmentation datasets. <strong>GS</strong>: Gleason Score. <strong>B</strong>: Benign. <strong>M</strong>: Malignant. <strong>#Slides (GS 3+3:3+4:4+3)</strong> Train Valid Test <strong>Total</strong> Biopsy 10:9:1 3:7:0 6:10:0 19:26:1 <strong>#Patches (B:M)</strong> Train Valid Test <strong>Total</strong> Biopsy 1557:2277 1216:1341 1543:2718 4316:6336 <strong>NB:</strong> Gland classification folder (gland_classification_dataset.zip) may contain extra patches, labels of which could not be identified from H&amp;E slides. They were not used in the machine learning study.

代码与*Patterns*论文链接如下：1. 多镜头神经机器翻译（Multi-lens Neural Machine, MLNM）代码 2. 用于低级别低体积前列腺癌高效诊断的AI辅助工具。

本研究从新加坡陈笃生医院的99名前列腺癌患者处，收集了40份前列腺切除术标本与59份穿刺针芯活检标本的数字化苏木精-伊红（H&E）染色全玻片图像（Whole Slide Image, WSI），共计99张全玻片图像，每份标本对应一张WSI。研究使用Leica Biosystems的Aperio AT2玻片扫描仪，以40×放大倍率对H&E染色玻片进行扫描（标本级像素尺寸为0.25μm × 0.25μm）。本研究已获得医院机构审查委员会批准，所有数据均完成去标识化处理。

穿刺针芯活检玻片中的前列腺腺体结构，使用ASAP标注工具（ASAP）进行手动标注与分类。一名资深病理医师对每张玻片内10%的标注结果进行复核，确保研究人员可获取活检不同区域的参考标注。需注意：出现在活检针芯边缘的部分腺体未被纳入标注范围。

**全玻片图像数据集**
本次提供的全玻片图像数据集（WSI.zip）包含99份SVS格式的图像，以及对应的XML格式标注文件。WSIs对应的患者分级信息可于`gleason_score_mapped.txt`中获取。上述XML标注可通过官方代码仓库中的代码进行解析。

**裁剪图像数据集**
本数据集包含从全玻片图像数据集的WSI中裁剪得到的512×512像素图像块，裁剪分辨率分别为5×、10×、20×与40×，且每张图像块均以标注腺体为中心。该数据集被用于训练两款AI模型，分别用于腺体分割（涵盖全部99名患者）与腺体分类（涵盖46名患者）。表1与表2分别展示了腺体分割与腺体分类数据集的相关信息。我们将两个对应子数据集打包为两个压缩文件，分别为：`gland_segmentation_dataset.zip`与`gland_classification_dataset.zip`。

**表1：** 腺体分割任务中训练集、验证集与测试集的玻片数量与图像块数量。每份前列腺切除术或穿刺针芯活检标本对应一张H&E染色WSI。

**#Slides（玻片数量）**
| 样本类型 | 训练集 | 验证集 | 测试集 | 总计 |
| ---- | ---- | ---- | ---- | ---- |
| 前列腺切除术标本 | 17 | 8 | 15 | 40 |
| 穿刺活检标本 | 26 | 13 | 20 | 59 |
| 总计 | 43 | 21 | 35 | 99 |

**#Patches（图像块数量）**
| 样本类型 | 训练集 | 验证集 | 测试集 | 总计 |
| ---- | ---- | ---- | ---- | ---- |
| 前列腺切除术标本 | 7795 | 3753 | 7224 | 18772 |
| 穿刺活检标本 | 5559 | 4028 | 5981 | 15568 |
| 总计 | 13354 | 7781 | 13205 | 34340 |

**表2：** 腺体分类任务中训练集、验证集与测试集的玻片数量与图像块数量。每份前列腺切除术或穿刺针芯活检标本对应一张H&E染色WSI。腺体分类数据集为腺体分割数据集的子集。
- **GS**：格里森评分（Gleason Score）
- **B**：良性（Benign）
- **M**：恶性（Malignant）

**#Slides（玻片数量，GS 3+3:3+4:4+3）**
| 样本类型 | 训练集 | 验证集 | 测试集 | 总计 |
| ---- | ---- | ---- | ---- | ---- |
| 穿刺活检标本 | 10:9:1 | 3:7:0 | 6:10:0 | 19:26:1 |

**#Patches（图像块数量，B:M）**
| 样本类型 | 训练集 | 验证集 | 测试集 | 总计 |
| ---- | ---- | ---- | ---- | ---- |
| 穿刺活检标本 | 1557:2277 | 1216:1341 | 1543:2718 | 4316:6336 |

**注：** 腺体分类数据集压缩包（`gland_classification_dataset.zip`）中可能包含额外的图像块，其标签无法通过H&E玻片识别，此类图像块未被纳入本机器学习研究中。