Knowledge Engine- A computational approach to combatting dermatological health disparities in underserved communities

摘要- 针对黑人女性销售的美发产品含有不成比例的有毒或过敏物质， 成分，研究表明，这些产品的化学暴露可能有助于健康 黑人女性的差距。为了说明这一点，一项对黑人妇女常用的18种主流产品的研究 发现100%的产品含有人造香料，已知会导致多种疾病，包括皮肤 刺激和湿疹，而72%的产品含有对羟基苯甲酸甲酯，一种与改变 孕妇的激素水平、雌激素和脂肪生成活性以及瘙痒风险增加 （皮肤瘙痒）在黑人妇女。与使用护发产品有关的其他不良健康结果包括接触性皮炎， 和脱发-影响估计50%的黑人妇女-以及乳腺癌的风险增加， 青春期过早、生殖损害和其他由生殖系统介导的疾病和状况。暴露于 外源性激素和内分泌干扰物（EDCs）也被报道为潜在的 导致月经提前、子宫肌瘤和不孕症。值得注意的是，黑人妇女的生殖 已经发现年龄越大，其体内的EDC水平越高（例如，对羟基苯甲酸酯、邻苯二甲酸酯）与白色 同龄的女人。据推测，在黑人妇女中观察到的高风险可能是由于 向他们销售的产品中较高浓度的内分泌干扰物和较高的应用频率，但 需要更多的研究来澄清潜在的联系，特别是因为黑色头发和皮肤在很大程度上被排除在外 用于产品开发和临床研究。满足进一步调查风险的需要 与发用产品成分、安全使用的最佳实践以及获取知识的适当途径相关 在社区内共享，HairDays提出了一种应用人工智能的多模式方法， (AI)包括机器学习（ML）和自然语言处理（NLP）在内的工具， 一种头发智能平台，将提供成分透明度，提供数据驱动的头发护理 建议，并提出具有文化意识的见解，以促进更安全的做法。产品 成分将使用从科学期刊和头发中挖掘的结构化和非结构化数据进行评估 护理/科学数据库，以及产品评论，以了解因素之间的关系， 成分、使用频率、头发轮廓和用户满意度。我们将利用这些知识成果， 对成分风险的见解和相应的产品建议，然后将传播到 受影响的社区通过HairDays平台。通过成功开发设想的 平台，我们的目标是减轻头皮，头发，和相关的病理（例如，潜在的乳腺癌风险） 不匹配的产品和高风险成分。第一阶段的具体目标如下：1）发展头发 使用NLP和AI的知识引擎提取和映射产品和成分数据;以及2）评估 用户对产品选择和使用的态度和信念。

Abstract - Hair products marketed towards Black women contain a disproportionate amount of toxic or allergenic ingredients, and research suggests that chemical exposure from these products may contribute to health disparities in Black women. Illustrating this, a study of 18 mainstream products commonly used by Black women found that 100% of products contained artificial fragrances known to cause multiple conditions including skin irritation and eczema, while 72% of products contained methylparaben, an ingredient associated with altered hormone levels in pregnant women, estrogenic and adipogenic activity, as well as an increased risk of pruritus (itchy skin) in Black women. Other poor health outcomes linked to hair product use include contact dermatitis, and hair loss—affecting an estimated 50% of Black women—as well as elevated risks for breast cancer, premature puberty, reproductive harm, and other hormone-mediated diseases and conditions. Exposure to exogenous hormones and endocrine-disrupting chemicals (EDCs) has also been reported as potentially contributing to cases of early menstruation, uterine fibroids, and infertility. Notably, Black women of reproductive age have been found to have higher levels EDCs in their bodies (e.g., parabens, phthalates) compared to white women of the same age. It has been postulated that the elevated risks observed among Black women may owe to both higher concentrations of EDCs in products marketed to them and higher frequencies of application, but more research is needed to clarify potential links, particularly as Black hair and skin have largely been excluded from R&D for product development and clinical research. Addressing the need for further investigation into risks associated with ingredients in hair products, best practices for safer use, and appropriate avenues for knowledge sharing within the community, HairDays proposes a multi-modal approach applying artificial intelligence (AI) tools including machine learning (ML) and natural language processing (NLP) to deliver a first-in- kind hair intelligence platform that will provide ingredient transparency, offer data-driven hair care recommendations, and present culturally conscious insights to promote safer practices. Product ingredients will be assessed using structured and unstructured data mined from scientific journals and hair care/science databases, as well as product reviews to understand the relationships between factors such as ingredients, frequency of use, hair profile, and user satisfaction. We will use these knowledge gains to generate insights on ingredient risk and corresponding product recommendations, which will then be disseminated to affected communities through the HairDays platform. Through successful development of the envisioned platform, we aim to mitigate scalp, hair, and related pathologies (e.g., potential breast cancer risk) stemming from ill-matched products and high-risk ingredients. Phase I Specific Aims are as follows: 1) Develop hair knowledge engine using NLP and AI to extract and map product and ingredient data; and 2) Assess impact of user engagement on attitudes and beliefs about product selection and usage.