Pulido-Arias Dagoberto, Henderson Rebecca, Millien Christophe, Lomil Joarly, Jose Marie Djenane, Flambert Gabriel, Bontemps Jean, Georges Emmanuel, Gunturi Alekhya, Shah Palak, Goncalves Tiago, Kalpathy-Cramer Jayashree, Gerstner Elizabeth, Wander Seth, Sirintrapun S Joe, Sgroi Dennis, Jeronimo Jose, Castle Philip E, Landgraf Kenneth, Brown Ali, Fadelu Temidayo, Shulman Lawrence N, Guttag John, Milner Dan, Brock Jane, Bridge Christopher, Kim Albert
Athinoula A. Martinos Center for Biomedical Imaging, Massachusetts General Hospital.
Department of Obstetrics & Gynecology, University of Alabama-Birmingham.
bioRxiv. 2025 Aug 28:2025.08.27.672012. doi: 10.1101/2025.08.27.672012.
Cancer morbidity disproportionately affects patients in low- and middle-income countries (LMICs), where timely and accurate tumor profiling is often nonexistent. Immunohistochemistry-based assessment of estrogen receptor (ER) status, a critical step to guide use of endocrine therapy (ET) in breast cancer, is often delayed or unavailable. As a result, ET is often prescribed empirically, leading to ineffective and toxic treatment for ER-negative patients. To address this unmet need, we developed (strogen Receptor tatus rediction for Haitian patients using deep learning-enabled histopathology hole Slide Imaging nalysis), a deep-learning (DL) model that predicts ER status directly from hematoxylin-and-eosin (H&E)-stained whole slide images (WSIs).
We curated two cohorts of H&E WSIs with tissue-matched ER status: The Cancer Genome Atlas (TCGA, n = 1085) and Zanmi Lasante (ZL, n = 3448) from Haiti. We trained two models using weakly supervised attention-based multiple instance learning: a "TCGA" model, trained on TCGA data, and ESPWA, trained on the ZL dataset. Model performance was evaluated using 10-fold cross validation.
Performance of the "TCGA" model was sensitive to the domain shift between the TCGA and ZL datasets, with a performance of an area under receiver operating characteristic (AUROC) of 0.846 on the TCGA test sets and 0.671 on the ZL test sets. Compared to the "TCGA" model, ESPWA demonstrated improved performance on the ZL cohort (AUROC=0.790; p=0.005). Subgroup analyses revealed clinically relevant populations in which ESPWA demonstrated improved performance relative to the overall cohort. Finally, ESPWA outperformed an academic breast pathologist (accuracy: 0.726 vs 0.639 respectively; p <0.001) in determining ER status from H&E WSIs.
ESPWA ("Hope" in Haitian Creole) offers an accessible framework to identify individualized therapeutic insights from H&E WSIs in LMICs. We have initiated clinical trials, using ESPWA, in ZL and sub-Saharan African countries to inform precision-based use of ET for prospective patients.
癌症发病率对低收入和中等收入国家（LMICs）的患者影响尤为严重，在这些国家，及时且准确的肿瘤分析往往并不存在。基于免疫组织化学对雌激素受体（ER）状态进行评估，这是指导乳腺癌内分泌治疗（ET）使用的关键步骤，却常常延迟或无法进行。因此，内分泌治疗常常是凭经验开出，导致对雌激素受体阴性患者的治疗无效且有毒副作用。为满足这一未被满足的需求，我们开发了（使用深度学习支持的组织病理学全切片成像分析预测海地患者雌激素受体状态），这是一种深度学习（DL）模型，可直接从苏木精-伊红（H&E）染色的全切片图像（WSIs）预测雌激素受体状态。
我们整理了两个具有组织匹配雌激素受体状态的H&E全切片图像队列：癌症基因组图谱（TCGA，n = 1085）和来自海地的赞米拉桑特（ZL，n = 3448）。我们使用基于弱监督注意力的多实例学习训练了两个模型：一个“TCGA”模型，在TCGA数据上进行训练，以及ESPWA，在ZL数据集上进行训练。使用10折交叉验证评估模型性能。
“TCGA”模型的性能对TCGA和ZL数据集之间的域转移敏感，在TCGA测试集上的受试者操作特征曲线下面积（AUROC）为0.846，在ZL测试集上为0.671。与“TCGA”模型相比，ESPWA在ZL队列上表现出更好的性能（AUROC = 0.790；p = 0.005）。亚组分析揭示了临床相关人群，其中ESPWA相对于总体队列表现出更好的性能。最后，在从H&E全切片图像确定雌激素受体状态方面，ESPWA优于一位学术乳腺病理学家（准确率分别为0.726和0.639；p <0.001）。
ESPWA（海地克里奥尔语中的“希望”）提供了一个可获取的框架，用于从低收入和中等收入国家的H&E全切片图像中识别个性化的治疗见解。我们已在ZL和撒哈拉以南非洲国家启动了使用ESPWA的临床试验，以为未来患者基于精准的内分泌治疗提供依据。
