2021-08-012024-05-18https://scholars.lib.ntu.edu.tw/handle/123456789/700103Certain limitations in in vitro fertilization (IVF) have remained insurmountable. The yield rate from IVF cycles remains low and live birth rates are reported to range from 20 to 30%. Poor embryo selection methodologies often necessitated the transfer of multiple embryos, which in turn increased the chances of multifetal gestations. Premature birth, underweight fetus, maternal eclampsia/pre-eclampsia, and gestational diabetes invariably result. Ovarian stimulation are also plagued by an inability to optimally predict treatment response, and inadequately stimulation cycles have resulted in failed oocyte retrieval or hyperstimulation cycles, which result in patient disappointment or significant patient morbidity. Therefore creating novel methods to improve the outcomes of artificial reproductive technologies should be important and in urgency. The integration of artificial intelligence (AI) and clinical medicine is definitely a promising trend. It can utilize bioinformatics and computerized technology to incorporate and analyze huge amount of complicated clinical data, including personal history, laboratory examination, imaging profiles, and even big data such as genomics, metabolomics, microbiota…etc. With the application of deep learning and machine learning technologies, a prediction model for the design and prognosis of IVF treatment can be established to achieve personalized precision medicine. Besides, so far there was no reliable non-invasive biomarker to predict the developmental potential of human embryos and therefore the success rate of embryo transfer is limited. Thus innovating biomarkers for embryo viability are necessary for clinical care and definitely have a huge commercialized potential. Therefore the aim of our proposal is to develop an interdisciplinary approach to overcome the current limitation of reproductive technologies. It will focus on four core objectives: (1) Improvement in oocyte fertilization rates by optimizing the selection of gametes; (2) Creation of an automatized embryo morphology assessment system to enhance the selection of the optimal embryo; (3) Identification of biomarkers that predict embryo implantation potential, and integrating this data with morphological analyses in order to improve clinical outcomes; (4) Creation of a predictive model for treatment responses in IVF, for patient-specific treatment. The combination of artificial intelligence and big data analysis with reproductive medicine is not only crucial and urgent, but an inevitable step in the progress of clinical and medical research.試管嬰兒技術經過近四十年發展，成功率至今仍十分有限，每進行一次試管嬰兒療 程的活產率僅約20-30%，也不乏排卵刺激不當而導致病患取卵失敗，或是出現卵巢 過度刺激併發症的案例。此外，由於現階段缺乏好的方法來挑選配子(精、卵)和胚 胎，因此臨床醫師必須一次植入多顆胚胎以增加懷孕率，反而增加病患多胞胎機率 ，並因此增加早產、胎兒體重不足、子癲前症、妊娠糖尿病等多種高危險妊娠的風 險。 隨著科技日新月異，臨床所能獲得的醫療資訊越來越多元而廣泛，人工智慧可運用 生物資訊及電腦科技技術，整合臨床病史、檢驗數據、影像資料、甚至基因體、代 謝體、微生物體……等眾多大數據以及複雜數位資訊，透過機器學習與深度學習進 一步建構完善的分析預測模型，不僅可為病患量身打造個人化精準醫療，提供更佳 的醫療服務，更能完善地蒐集、儲存並整合海量複雜的臨床資料，提供將來醫學研 究發展極佳的資源。儘管人工智慧在生殖醫學的相關應用極具發展潛力，然而截至 目前為止無論是醫學研究或是實際臨床應用都十分闕如，雖有少數私人企業透過機 器學習技術研發胚胎篩選影像分析系統，但實際成效缺乏驗證，也沒有足夠良好的 非侵入性生物指標可以作為篩選良好胚胎的標準。 因此本計畫預計將人工智慧醫療以及大數據分析整合應用於生殖醫學，將規劃四大 核心目標：(1) 運用微流體晶片或其他創新技術提升精蟲篩選能力以促進卵子受精 率；(2) 創建胚胎影像分析系統，提升篩選胚胎品質的能力，精準挑選具發育潛力 的胚胎進行植入，增加成功率並減少多胞胎併發症；(3) 尋找可預測胚胎著床能力 之生物標記，包括近年來極為熱門的微小核醣核酸研究，並結合圖像資料的人工智 慧判讀以及生物標記的臨床運用，開發新的胚胎篩選機制增加試管嬰兒治療懷孕率 ，並降低多胞胎妊娠機率；(4) 創建試管嬰兒療程預測模組以優化整個療程進行 ，並研究新的試管嬰兒療程預測指標，例如不孕病患之微生物體分布表徵與成功率 之相關聯性，依照病患特色規劃最適當排卵刺激療程，以增加卵子成熟率、胚胎受 精率、胚胎著床率，並減少不必要的醫療花費以及潛在副作用，達到精準醫療的最 終目標。總結來說，人工智慧及大數據分析與生殖醫學的結合，無論是在臨床服務 或者醫學研究都十分關鍵重要且迫在眉睫。