Research Impact This project addresses the growing prevalence of neurocognitive disorders (NCDs) in an aging global population, with the 60+ demographic set to double by 2050. In Hong Kong, over 100,000 older adults may go undiagnosed due to long wait times and high care costs. We develop AI-driven spoken language analysis technologies for automated, non-intrusive NCD screening, offering cost-effective, scalable, and accessible cognitive assessments. With the CU-MARVEL corpus, the largest for NCD research in Hong Kong, we advanced deep learning for spoken language biomarkers. Integrating speaker diarization, speech recognition, and NLP, we created an AI-enabled screening pipeline with strong results in both English and Cantonese. Novel tasks like the Hong Kong Grocery Shopping Dialog and fMRI movie-watching have improved screening via memory and communication. Our screening apps are being piloted in the community. This work aligns with important initiatives such as Healthy China 2030 Blueprint, WHO’s Global Action Plan on Dementia, and the UN’s 2030 Agenda for Sustainable Development by enhancing healthcare access and reducing inequalities. 研究影響 這個計劃應對了全球人口老化過程中神經認知障礙（NCDs）日 益增長的嚴重問題，預計到 2050 年，60 歲以上人口將翻倍。 在香港，由於漫長的候診時間和高昂的醫療費用，超過 10 萬名 老年人可能無法及時確診。我們開發了基於 AI 的語言分析技術， 用於自動化、非侵入式的 NCD 篩檢，提供具有成本效益、可擴 展且易於取得的認知評估。透過香港最大的 NCD 研究語料庫 CU-MARVEL，我們推動了語言生物標記的深度學習研究。透過 整合說話者分離、語音辨識和自然語言處理（NLP），我們創建了 一個支援 AI 的篩檢管道，在英語和粵語中均取得了出色成果。像 是香港購物對話任務和 fMRI 觀影任務這樣的新穎任務，改善了 對記憶和溝通的篩選。我們的篩檢應用程式目前正在社區中進行 試點。該工作與《健康中國 2030 藍圖》、世界衛生組織的《腦退 化全球行動計劃》和聯合國《2030 年永續發展議程》等重要倡 議保持一致，旨在提升醫療服務的可及性並減少不平等現象。 Abstract Population ageing is a global issue. The WHO projects that by 2050, 22% of the world's population will be aged 60 or older, with Hong Kong's population aged 65+ rising to 35%. Population ageing is associated with high-burden geriatric syndromes, increasing public healthcare costs and threatening societal sustainability due to a shrinking workforce and tax base. Neurocognitive disorders (NCD), including dementia, are especially prevalent among older adults, with care costs estimated at USD 1 trillion today, expected to double by 2030. Effective disease screening and management are crucial. Current NCD diagnoses rely on clinical professionals using neuropsychological tests, which are limited by clinician shortages, subjective assessments, and cultural biases. To address these issues, an automated, objective evaluation platform will be developed using inexpensive spoken language biomarkers for NCD screening and monitoring. This platform will enable remote monitoring, generating patient alerts for timely treatment. Collecting individualized data over time will help detect early cognitive decline, improving disease management and reducing care costs. Spoken language biomarkers will be targeted, as they are non-intrusive and can be easily captured remotely. AI-driven technologies will extract these biomarkers, providing sensitive cognitive assessments. This research aligns with WHO’s goals and aims to support patients and caregivers in Hong Kong through AI-enabled healthcare. 項目簡介 人口老化是一個全球性問題。根據世界衛生組織預測，到 2050 年，全球 60 歲以上人口將佔 22%，而香港 65 歲以上人口將上升 至 35%。老化伴隨著高負擔的老年綜合症，增加了公共醫療費 用，並由於勞動力和稅基的縮小而威脅到社會的可持續性。神經 認知障礙（NCD），包括失智症，在老年人中尤其普遍，目前護 理成本估計為 1 兆美元，預計到 2030 年將翻倍。有效的疾病篩 檢和管理至關重要。目前的 NCD 診斷依賴臨床專業人員使用神 經心理測試，受限於醫生短缺、主觀評估和文化偏見。為解決這 些問題，本項目將開發一個基於廉價生物標記的 自動化、客觀評 估平台，用於 NCD 篩檢和監測。該平台將實現遠端監測，產生 患者警報以便及時治療。透過長期收集個人化數據，有助於早期 檢測認知衰退，提高疾病管理水平，並降低照護成本。口語語言 生物標誌物會成爲提取目標，因為它們具有非侵入且易於遠端捕 獲的特性。透過 AI 技術提取這些生物標誌物，提供敏感的認知評 估。這項研究與世界衛生組織的目標一致，旨在透過 AI 支持的醫 療保健幫助香港的患者和照護者。 The inter-disciplinary and inter-university team for the project: (front-left) Professor XM Gong (CU), Professor XX Wu (CU), Professor Thomas Lam (CU), Professor Bonnie Lam (CU), Professor Brian Mak (HKUST), Professor MW Mak (PolyU), Professor Helen Meng (CU), Professor Vincent Mok (CU), Professor Helene Fung (CU), Professor Timothy Kwok (CU), Professor Diana Lee (CU), Professor XJ Ma (HKUST), Professor Kelvin Tsoi (CU), Professor Andrew Liu (CU, absent), Professor Patrick Wong (CU, absent) and Professor Jean Woo (CU, absent); (upper) students, postdoctoral, research assistants and supporting staffs. 跨學科及跨大學的計劃團隊︰( 前排左起 ) 龔先旻教授 ( 中大 )、吳鍚欣教 授 ( 中大 )、林遠東教授 ( 中大 )、林賢嘉教授 ( 中大 )、麥鑑榮教授 ( 科大 )、 麥文偉教授 ( 理大 )、蒙美玲教授 ( 中大 )、莫仲棠教授 ( 中大 )、馮海嵐教 授 ( 中大 )、郭志銳教授 ( 中大 )、李子芬教授 ( 中大 )、麻曉娟教授 ( 科 大 )、蔡錦輝教授 ( 中大 )、劉循英教授 ( 中大、缺席 )、黃俊文教授 (中 大、缺席 )、胡令芳教授 ( 中大、缺席 )；( 後排 ) 研究生、博士後、研究 人員及支援職員 Demonstration for the project’s app in the CUHK Medical Centre Innovation Week (December 15, 2023) 在香港中文大學醫院創新週活動中公開展示計劃的應用程式 Awarded Silver Medal in the 49th International Exhibition of Inventions Geneva 2024 榮獲 2024 年第 49 屆日內瓦國際發明展銀獎 25
RkJQdWJsaXNoZXIy NDk5Njg=