SHIDC-BC-Ki-67

核蛋白 Ki-67 和肿瘤浸润淋巴细胞 (TIL) 已被引入作为预测肿瘤进展和对化疗可能反应的预后因素。 Ki-67 指数和 TIL 在治疗异质性肿瘤中的价值，如乳腺癌 (BC)，这是全世界女性最常见的癌症，已在文献中得到强调。考虑到对这两个因素的估计都依赖于专业病理学家的观察，并且也可能存在个体间的差异，使用机器学习的自动化方法，特别是基于深度学习的方法，已经引起了人们的关注。然而，深度学习方法需要大量注释数据。在没有公开可用的 BC Ki-67 细胞检测基准和进一步注释的细胞分类的情况下，在本研究中，我们提出 SHIDC-BC-Ki-67 作为上述目的的数据集。我们还引入了一种新的管道和后端，用于估计 Ki-67 表达和同时确定乳腺癌细胞中的肿瘤内 TIL 评分。此外，我们表明，尽管我们提出的模型遇到了挑战，但我们提出的后端 PathoNet 在获得的调和平均测量方面优于迄今为止提出的最先进的方法。

Nuclear protein Ki-67 and tumor-infiltrating lymphocytes (TIL) have been established as prognostic factors for predicting tumor progression and potential response to chemotherapy. The prognostic value of Ki-67 index and TIL in heterogeneous tumors, including breast cancer (BC) — the most prevalent malignancy in women globally — has been extensively emphasized in existing literature. Given that the estimation of both factors relies on observations from specialized pathologists and may suffer from inter-observer variability, automated machine learning-based approaches, particularly deep learning-based methods, have garnered significant research attention. However, deep learning approaches require large volumes of annotated data. In the absence of publicly available benchmarks for BC Ki-67 cell detection and annotated cell classification, we propose SHIDC-BC-Ki-67 as a dedicated dataset for the aforementioned purposes in this study. We also introduce a novel pipeline and backend framework for estimating Ki-67 expression and concurrently determining intratumoral TIL scores in breast cancer cells. Furthermore, we demonstrate that despite the challenges faced by our proposed model, our introduced backend PathoNet outperforms the current state-of-the-art methods in terms of the obtained harmonic mean metric.

搜集汇总

背景与挑战

背景概述

SHIDC-BC-Ki-67是一个专注于乳腺癌的基准数据集，旨在提供Ki-67核蛋白和肿瘤浸润淋巴细胞（TIL）的注释数据，以支持基于深度学习的自动化检测方法。该数据集填补了公开可用数据空白，并引入了一种新管道和后端PathoNet，用于同时估计Ki-67表达和TIL评分，在性能上优于现有先进方法。