Abstract

Cofilin-1 is a member of the actin depolymerizing factor (ADF)/cofilin family. Previous research in our laboratory has shown that overexpression of cofilin-1 leads to cell cycle arrest in G1 phase and induces senescence. However, the effects of cofilin-1 expression on mammalian aging remain unclear. Therefore, we used a Cre/loxP system transgenic mouse model to increase the expression of cofilin-1 and investigate whether the overexpression of cofilin-1 led to aging and the appearance of aging features.

In this study, we initially confirmed the aging phenotype in these transgenic mice through behavioral analysis, protein expression studies, and biomedical imaging. Our results revealed that cofilin-1 overexpression led to reduced activity levels, impaired motor coordination, and memory deficits, as well as increased expression of aging-related proteins and structural changes observed through MRI and PET imaging. These findings suggest that cofilin-1 overexpression induces premature aging characteristics in this transgenic mouse model.

We believe that this animal model can serve as a rapid and effective system for studying aging-related mechanisms and can potentially be used to accelerate the development of aging-related research.

本團隊研究發現ADAM9調控KRAS的新機轉，研發出ADAM9抑制劑，具可作用於多種KRAS突變型的優勢，故能廣泛應用於胰臟癌治療。藥物具專一性、成藥性佳、有效縮小腫瘤、抑制癌轉移。ADAM9 抑制劑合併化療藥 Gemcitabine，可強化Gemcitabine的療效，具顯著的藥效加乘性，並兼具治療安全性，達成根除腫瘤的最佳目標。此研究已發表於2024 Nature Cancer。

Our team discovered a new mechanism by which ADAM9 promotes KRAS activation. We developed ADAM9 inhibitors, which function as pan-KRAS inhibitors and widely reduce KRAS proteins in pancreatic cancers. The drug is specific with good drug properties, effectively reducing tumors, and inhibiting cancer metastasis. ADAM9 inhibitors can enhance the efficacy of Gemcitabine to eradicate tumors in pancreatic cancer treatment. The study has been published in Nature Cancer 2024.

糖尿病患者的眼睛會「提早老化」。因為血糖過高會使眼部神經的血液循環跟水晶體代謝變差，罹患白內障與青光眼的機率也較高。因此若能降低高糖對眼睛的傷害，將可以造福糖尿病族群，如能開發適當的治療藥物將可降低白內障與青光眼的風險。但由於眼睛結構特殊性，所以藥物不易穿透角膜來降低治療或是延緩糖尿病白內障發生，因此開發相關治療藥物是目前所需解決問題。
本研究將以目前臨床使用sodium glucose cotransporter 2 (SGLT2)血糖運輸阻斷劑抑制DM患者白內障之惡化速度。為了提升Dapa之治療效果，本研究經由設計奈米載體裝載Dapa形成眼藥水配方，該載體之生物功能將能克服Dapa水溶性/淚液代謝/結膜及角膜阻礙等影響藥物療效之生物屏障。就經濟影響而言，本研究開發之載體藥物配方，不僅能應用於抑制糖尿病患者白內障惡化之速度，更可應用於提升其他眼科用藥之治療效果。

有效的抗癌療法開發於臨床前階段面臨許多阻礙，包含缺乏具有人體生理代表性的測試平台、缺乏針對腫瘤環境綜合分析之技術、與操作複雜度高而難以自動化。因此，我們提出了MedSelect技術。鑒於以上未被滿足需求，MedSelect系統整合患者來源之三維腫瘤組織培養、與動態培養基循環技術，能夠重建組織微環境中的動態分子和細胞遷移過程，並具有生理環境下的分子梯度，同時由於關鍵專利設計MedSelect能夠實時分析不同異質性環境下的腫瘤反應。最後，MedSelect系統整合三維腫瘤建模、培養與分析的自動化，以解決現有方法人力負擔高之限制。

我們的藥物創新，源於發現一個Aβ42在代謝調節過程的新穎生理功能，Aβ42透過與一個轉錄因子的結合，調控葡萄糖和脂質平衡、能量消耗和胰島素敏感性的基因表達。藥物設計針對這一轉錄因子的作用機制；我們篩選了多種FDA批准的藥物組合，以恢復正常代謝功能和免疫反應為目標。在臨床前動物的研究，我們發現藥物有效減輕了阿茲海默症(AD)小鼠的代謝與免疫異常，並改善腦病理變化。更重要的是，我們利用台灣健保資料庫進行回顧性分析，發現使用這種組合藥物可降低AD發病率超過55%。這些令人鼓舞的發現顯示這是一個早期干預AD的有力候選藥物，可以直接進入臨床試驗。目前，我們正在開發一種貼片，以將這種組合藥物直接遞送到特定組織。這種遞送方法旨在降低治療效果所需的劑量，同時將此老藥的新療法與現有藥物的非標籤使用區分開來，從而提升其市場吸引力和商業成功。
Our drug innovation arises from identifying a novel cellular role of Aβ42 in metabolic modulation through its interaction with and regulation of a transcription factor involved in regulation of the expression of genes involved in glucose and lipid homeostasis, energy expenditure, and insulin sensitivity. The drug design targets this underlying mechanism. We screened various combinations of FDA-approved drugs to restore both dysregulated metabolic functions and immune responses in the early stages of Alzheimer’s disease (AD). In preclinical studies, our drug effectively mitigated brain pathology in early AD mouse models. Remarkably, a retrospective population-based analysis revealed that the use of this combination drug was associated with a more than 55% reduction in the incidence of dementia. These promising findings suggest that our combination drug is a strong candidate for direct entry into clinical trials for early intervention in AD. Currently, we are developing a patch to deliver this combination drug directly to specific tissues. This delivery method aims to lower the required doses needed to achieve therapeutic effects while also differentiating the new therapy from the off-label use of existing drugs, thereby enhancing its market appeal and commercial success.

在當今醫學中，將藥物精確地遞送到目標部位仍然是一項重大挑戰。許多藥物在臨床試驗中失敗的原因在於，現有的藥物傳遞仍依賴高滲透性和滯留（enhanced permeability and retention effect，EPR）效應來到達目標部位，導致大部分藥物無法成功地運輸到病變部位。然而，許多臨床數據顯示EPR效應只能略微提高疾病組織中的藥物可用性。因此，許多藥物並非因為缺乏療效，而是因為無法有效地遞送到需要治療的部位。針對此一問題，本團隊設計一種新型藥物載體，無需依賴EPR效應即可將藥物精準遞送到病變部位，顯著提高藥物在病變組織中的可用性。本團隊的創新技術為克服藥物遞送挑戰提供了突破性的解決方案，將有效改進臨床治療效果。
單核細募集是許多慢性疾病常見的病理現象，包括癌症、神經退化性疾病和心血管疾病。一旦進入循環系統，這些循環中的單核細胞對腫瘤和缺血組織等異常組織具有強大的定向能力，能夠有效地到達如腫瘤和缺血組織等病變區域。針對這一病理現象，本團隊開發了一種創新的藥物載體，可以搭載到這些循環中的單核細胞上，利用這些細胞作為「專車司機」，將藥物精準運輸到目標部位。本團隊將此一藥物載體命名為為單核細胞介導的藥物載體（monocyte-mediated drug carriers，MMDCs）。藉由3D微流道體外腫瘤模型和人類乳腺腫瘤移植小鼠模型，本團隊證明至少有40%以上被注射的MMDCs能成功搭載到循環中的單核細胞上。此外，MMDCs與健康內皮細胞的相互作用極低，顯示這些藥物載體不會引起不良的凝血效應。在搭載到細胞表面後，MMDCs將穩定地停留在循環單核細胞的表面，而不會立即被細胞吞噬，從而防止藥物在循環過程中的過早釋放。
一旦進入腫瘤組織，帶有MMDCs的單核細胞會分化成巨噬細胞，並將原本貼附在細胞膜上的MMDCs吞噬進細胞內。被吞食的MMDCs其結構會被破壞，讓被包覆的藥物成功的釋放出來。本團隊發現在腫瘤組織中MMDCs的數量顯著高於目前臨床常使用的聚乙二醇化脂質體。此外，MMDCs包覆的阿黴素治療組的小鼠，其腫瘤明顯小於接受聚乙二醇化質體包覆的阿黴素（LipoDOX）治療組的小鼠。本團隊研發的創新脂質體製劑具備高效遞送化療藥物的能力，更令人矚目的是其多功能性，不僅能有效輸送組織再生藥物及用於包覆顯影劑，從而協助醫生精確辨識腫瘤和缺血組織等異常部位的精確位置。藉由MMDCs包覆顯影劑，與先進的人工智慧技術相融合，我們更能夠精密建立高度精準的三維病灶模型，更能為個人化精準醫學治療提供強有力的支持及開啟一扇嶄新之門，開啟醫療科技的全新篇章，引領人類健康照護的未來方向。

癌症是全球頭號死亡原因，在台灣尤為嚴重。因此，抗癌藥物的研發刻不容緩。化療是目前癌症治療的主要手段之一，其中核酸標靶藥物佔比超過50%。然而，現有的核酸標靶抗癌藥物有副作用大、抗藥性高、特異性低等缺點。為了解決這些問題，我們建立了一個獨特的藥物發現平台，用於開發精確靶向癌症相關 DNA 的化療藥物。這個平台包括我們自己創建的兩個資料庫，以及AI輔助的藥物篩選和設計技術，以增強化療藥物的特異性並減少副作用。透過這項技術，我們開發了一系列新型化合物（NA化合物），它們在與錯配鹼基配對的DNA結合方面表現出高度特異性，而這些化合物已被證明在細胞和動物模型中可有效對抗錯配相關的大腸癌，且具有低副作用的表現。

本研發產品致力於利用環型核醣核酸（circular RNA）的高穩定性，解決RNA冷鏈運輸需求，並大幅提升細胞內的持久表現。我們開發的『環型核醣核酸蛋白質增益表現』(EPIC)系統融合了獨特的轉譯控制技術，具備高效率的背向剪接控制模組，顯著增強蛋白質的轉譯活性，並最大化治療性蛋白質的表現潛力。傳統上，環型核醣核酸被認為是microRNA的吸附分子，但我們的技術突破其限制，有系統地削弱此作用。此外，EPIC可讓circRNA專一留存於細胞質中，促進長期穩定的蛋白質。透過這項技術，我們為基因編輯、免疫治療和疫苗開發等臨床應用提供了更持久且有效的解決方案，展現出極具市場潛力的創新方向。這項產品將為核酸藥物的臨床應用提供嶄新的替代選擇。

腫瘤中存在含量極少卻造成腫瘤抗藥性、復發、轉移的癌幹細胞，本團隊所開發之技術相較傳統分選方法能篩選與培養出6倍以上穩定度(天數)、10倍以上基因表現量與50倍以上數量的癌幹細胞庫，未來將藉由本計畫衍生新創公司於全球販售癌幹細胞株及專屬培養液，並提供客製化癌幹細胞株建置之服務，運用癌幹細胞抗原建置標靶癌幹細胞之細胞療法並申請特管法治療以及相關可能新藥開發等，以有效解決當前臨床治療在腫瘤復發及轉移所遇到之困境。

現有阿茲海默症治療藥物療效不佳、副作用明顯，亟需研發新一代藥物。NMDA受體功能低下為阿茲海默症的機轉之一，活化NMDA受體可以增加認知功能。D-amino acid oxidase (DAAO)負責分解NMDA受體神經傳導物質D-serine 等內生性胺基酸，苯甲酸鈉是一種DAAO抑制劑，並具抗氧化功效。抑制DAAO可增強NMDA受體的神經傳導。研究團隊的臨床試驗發現苯甲酸鈉可改善阿茲海默症病患的認知與整體功能，無明顯副作用，已獲得台灣、澳洲、加拿大、德國、日本、南韓、新加坡等七國專利、並完成技術移轉至科進製藥股份有限公司，於通過TFDA和各醫院IRB的審查後，已在台大、林口長庚、雙和、高醫四家醫院進行二期臨床試驗。苯甲酸鈉的研發預期可成為新一代阿茲海默症治療藥物，為全球廣大的失智症病患帶來福祉。

This project aims to "develop an oral vaccine for avian influenza in poultry" to combat and prevent the spread of avian influenza among birds, avoid economic losses in the poultry industry, and even prevent cross-species transmission to humans.

背景：腦炎是中樞神經系統 (CNS) 的發炎症疾病，與腦功能障礙的臨床證據有關。 它可能由細菌、病毒或真菌感染引起。 許多病毒，如腸病毒、皰疹病毒和流感病毒，都可能導致這種神經系統疾病，其中又以小RNA病毒科（Picornaviridae）的感染為全世界大多數病毒性腦膜炎病例的致病原，然而腦炎的發病機制仍不清楚，需要更多的研究來幫助臨床診斷和治療。Human parechovirus type 3 (PeV-A3) 是一新興病毒傳染病，PeV-A3 被認為是為造成嬰幼兒腦膜腦炎主要病原之一。目前的研究PeV-A3病毒性腦炎的致病機制與藥物研發以2D細胞株為主，然而在細胞株平台所發現的致病機轉與篩選出的藥物的治療效果，時常無法在實驗動物或人體上再現。再者，我們之前的研究顯示，實驗小鼠並不是一個適合研究PeV-A3 病毒與宿主間的交互作用的感染模式，因為PeV-A3 病毒是一種僅感染人類的小RAN病毒，無法有效的在實驗小鼠建立一個高致病性的感染，此不同物種間的病毒感受性與生理差異限制了我們對人類病毒的致病機制的通盤了解。因此，我們需要一種介於細胞株與實驗動物之間的一種新的病毒感染研究模式，以了解病毒性腦炎的致病機轉並發展治療藥物。
方法：利用幹細胞 (stem cells) 發育為人類3D類器官的培養技術是生物醫學領域的一個重要里程碑，並已應用於多種研究領域。然而使用人類3D類器官進行病毒學的研究與應用仍為少見，急需積極開發。而利用誘導性多功能幹細胞 (induced pluripotent stem cells, iPSC) 發育為人類 3D 大腦類器官是一個十分有潛力的研究模式，除了iPSC細胞取得容易，亦能避免使用胚胎幹細胞所衍生的倫理議題，因此很適合用來發育為人類 3D 大腦類器官來探討 PeV-A3 腦膜腦炎並建立抗病毒藥物研發平台。
結果與展望：我們已成功利用iPSC 成功培養出人類 3D 大腦類器官，並以細胞學與組織學分析證明，我們的人類大腦類器官具有人類大腦的各組成細胞。我們再利用 PeV-A3 感染此發育的大腦類器官，並以免疫螢光染色偵測病毒，證明 PeV-A3 能順利感染 3D 大腦類器官，且在抗病毒激素-干擾素的存在下，PeV-A3 感染受到抑制，此系統的建立可以於發展治療 PeV-A3 病毒性腦膜腦炎將有極大的幫助。因人類 3D 大腦類器官相對於需要維持動物照護設施進行動物實驗的技術門檻與經費需求較低，因此非常適合進行商化技轉，並擴大應用於在神經系相關疾病。
Background: Encephalitis is an inflammatory disease in the central nervous system (CNS) that is associated with clinical evidence of brain dysfunction. It can be caused by bacterial, viral, or fungal infections. Many viruses, such as enteroviruses, herpesviruses, and influenza viruses, can lead to this neurological disease, with infections from the Picornaviridae being the causative agents in most cases of viral meningitis worldwide. However, the pathogenesis of encephalitis is still unclear and further research is needed to aid in clinical diagnosis and treatment. Human parechovirus type 3 (PeV-A3) is an emerging viral infectious disease and is considered one of the main pathogens causing infantile meningitis. Current research on the pathogenic mechanisms and drug development for PeV-A3 viral encephalitis primarily relies on 2D cell cultures. However, the pathogenic mechanisms discovered in cell culture platforms and the therapeutic effects of screened drugs often fail to be reproduced in experimental animals or humans. Furthermore, our previous studies have shown that experimental mice are not a suitable infection model for studying the interactions between PeV-A3 virus and the host because PeV-A3 only infects humans, and it cannot establish a highly pathogenic infection in experimental mice. The differences in viral susceptibility and physiological variations between different species limit our comprehensive understanding of the pathogenic mechanisms of human viruses. Therefore, we need a new virus infection research model that bridges the gap between cell cultures and experimental animals to understand the pathogenesis of viral encephalitis and develop therapeutic drugs.
Methods: The cultivation technique of developing human 3D organoids from stem cells is a significant milestone in the field of biomedical research and has been applied in various research areas. However, the use of human 3D organoids for virology research and applications is still relatively rare and urgently needs active development. Developing human 3D brain organoids from induced pluripotent stem cells (iPSC) is a highly promising research model. Besides the ease of obtaining iPSC cells, it also avoids the ethical issues associated with the use of embryonic stem cells, making it suitable for developing human 3D brain organoids to investigate PeV-A3 meningitis and establish a platform for antiviral drug development.
Results and Perspective: We have successfully cultivated human 3D brain organoids using iPSC and demonstrated through cellular and histological analyses that our human brain organoids possess various constituent cells found in the human brain. We then infected these developed brain organoids with PeV-A3 and detected the virus using immunofluorescence staining, confirming that PeV-A3 can effectively infect the 3D brain organoids. Furthermore, in the presence of antiviral hormones-interferons, PeV-A3 infection was inhibited. The establishment of this system will greatly aid in the development of treatment for PeV-A3 viral meningitis. Compared to animal experiments that require maintaining animal care facilities, human 3D brain organoids have lower technical and financial requirements, making them highly suitable for commercialization and broader application in neurologically related diseases.

Needs：(1)Patient who is Struggling with lowfrequency noise 24 hours a day, 7 days a week by the machine. (2) How to have a good Life quality?

Proposed solution：Current market solutions for small and thin sound-absorbing materials with reducing low-frequency noise.

Echo-free: Our material achieves no echo at 722 Hz with a thickness of only 2.5 cm.

使用3D生物列印技術製造含血管類骨器官(VBO)，模仿骨髓腔內骨小樑和骨內膜，在中空立方形網狀聚己內酯(PCL)-骨支架上培養人類骨前驅細胞，並模仿骨髓腔血管網路，列印混合人類內皮前驅細胞之丙烯酸甲酯凝膠(GelMA)-血管支架，將其置入兩片骨支架之間，上述兩者共同培養形成VBO。將來可以使用患者自己的骨髓細胞，配合最新發展之支架材料，及最佳機械壓縮力和灌注流體力參數條件的刺激下，在短時間(例如一個月內)將VBO培養成為自體骨移植物的最佳選擇。

We have developed a novel 3D bioprinting strategy to fabricate human pre-vascularized bone organoids (VBO). The process involves creating hollow cubic bone scaffolds made of polycaprolactone (PCL) with osteoprogenitor cells. VBOs are vascularized by inserting vessel-like structures formed by bioprinting Gelatin-methacryloyl (GelMA) hydrogel loaded with human umbilical vein endothelial cells (HUVECs) between two bone organoids. A compression-fluidic perfusion system provides mechanical forces to enhance VBO toughness and vascularization. This system includes electric gripper chambers that hold and detach various VBO designs, with medium perfusion flow managed by a peristaltic pump and uniaxial compression by a stepper motor-driven gripper. VBO development combines computer science, materials, machinery, and biomedicine, creating a platform for drug development for osteoporosis, improving bone fracture healing, offering an alternative to autologous bone grafts, and reducing the need for animal experimentation in skeletal disease studies.

自動化尿液處理系統：
長時臥床或重症病患常使用導尿管，協助尿液排出並監測生理功能。然而，尿袋處理對照護者而言是沉重的負擔。護理人員需定時倒尿、紀錄尿量、顏色，並清潔尿袋口，不僅耗時耗力，也潛藏感染風險。
為了解決此問題，我們開發出「自動化多功能尿液處理系統」。此系統可自動倒尿、測量尿液重量、辨別顏色、處理廢棄尿液，並整合醫院資訊系統，大幅減輕照護負擔、提升醫療效率。
隨著人口老化、醫療發展、以及獨立病房需求增加，導尿管和照護機器人的市場規模不斷擴大，自動化尿液處理系統具有廣泛的剛性需求，潛力十足。

Long-term bedridden or critically ill patients often rely on indwelling catheters to facilitate urine drainage and monitor physiological functions. However, managing urine bags poses a significant burden on caregivers. Nursing staff must regularly empty urine bags, record urine volume and color, and clean the catheter ports, a process that is not only time-consuming and labor-intensive but also carries the risk of infection.

To address this challenge, we have developed an ""Automated Multi-Task Urine Management System."" This system automates urine emptying, measures urine weight, identifies urine color, processes waste urine, and integrates with hospital information systems, significantly reducing caregiver burden and improving healthcare efficiency.

Project Content Summary: Addressing common bleeding situations in general surgery, this project aims to utilize silk protein as raw material. Through ultrasonic chemical and electrochemical processes, nanoscale silk protein microfiber hemostatic powder will be developed. This powder can rapidly and effectively stop bleeding, reducing surgical risks, minimizing blood transfusion volume, shortening postoperative drainage time, and lowering infection risks.
Expected Outcomes: The team will collaborate with manufacturers to prioritize the development of nanoscale microfiber hemostatic powder using medical-grade silk protein. In the second phase, they will combine this nanosilk protein hemostatic powder with powder sprayers or endoscopic instruments to create advanced hemostatic medical devices with clinical convenience.
Industry Benefits: The global hemostatic dressing market value is projected to exceed $5.4 billion by 2025, with a compound annual growth rate (CAGR) of 5.3% from 2020 to 2025. The primary markets are the United States and the European Union. Currently, there are 29 advanced hemostatic medical devices domestically, all relying on imports. The successful execution of this project aims to establish domestically produced high-end medical products.

全球軟骨修復市場規模預計2032年達到約46.2億美元，2023~2032年複合年增長率為12.1%；隨著細胞治療法規的完善，組織工程再生醫療產品將成為未來有效治療關節軟骨缺損的趨勢。
我們開發新型3D複合材料，使用去細胞豬軟骨粉做為支架，添加高濃度血小板血漿和凝血酶作為生長因子，結合自體軟骨細胞形成用於治療膝關節軟骨缺損的3D工程軟骨植入物。在體外和體內實驗中，此複合植入物能促進軟骨標誌物-Collagen II與Aggrecan的表達，並在全層軟骨缺損中再生透明軟骨。我們積極將此3D工程軟骨植入物推向臨床，以改善目前難以治癒的關節軟骨缺損。


The global cartilage repair market is expected to reach approximately USD 4.62 billion by 2032, with a compound annual growth rate (CAGR) of 12.1% from 2023 to 2032. This indicates growing interest and investment in advanced treatments, particularly in cell-engineered regenerative medicine. As regulations for cell therapy become increasingly refined, cell-engineered regenerative products are anticipated to become the future trend for effectively treating articular cartilage defects.
We have developed a novel 3D composite material using decellularized porcine cartilage graft as a scaffold. Platelet-rich plasma and thrombin are added as growth factors, combined with autologous chondrocytes to form a 3D engineered cartilage implant for treating knee joint cartilage defects. In both in vitro and in vivo experiments, this composite implant can promote the expression of cartilage markers-Collagen II and Aggrecan, and regenerate hyaline cartilage in full-thickness cartilage defects. We are actively advancing this 3D engineered cartilage implants toward clinical application to improve the currently difficult-to-heal articular cartilage defects.

成功大學的 VRMT 團隊致力於推動我們的虛擬實境鏡像治療（VRMT）系統，來促成中風復健產業的革新。通過將沉浸式虛擬實境與傳統鏡像治療相結合，我們的創新系統不僅增強了患者上肢運動功能，還加速了中風患者的恢復。VRMT系統已完成FDA二級醫材註冊，有嚴謹的研究和臨床試驗的支持，致力於在中風康復領域樹立新的標準，改善患者的護理品質，提升醫療效率。
The NCKU VRMT team is pioneering in advancing stroke rehabilitation through our Virtual Reality Mirror Therapy (VRMT) systems. By blending immersive virtual reality with conventional mirror therapy, our innovative system enhances motor function and accelerates recovery for stroke survivors. Our approach offers a compelling rehabilitation experience, backed by rigorous research and clinical trials. VRMT has finished the registration of FDA through Class II exempt. Our multidisciplinary team is committed to setting new standards in stroke recovery and improving patient care.

表面增強拉曼光譜（SERS）能夠增強拉曼訊號，作為精確識別分子的分子條碼，儘管信號微弱且存在螢光干擾。我們開發了一種綜合SERS檢測平台（CSDP），包含具三維電晶熱點之訊號增強晶片和信號處理軟體，在無需樣品預處理的情況下將拉曼訊號增強了1000倍以上，僅需極少的樣品量和短暫的檢測時間。該晶片利用等離子共振和交叉銀納米線的均勻熱點分佈，提供儲存穩定性，並兼容任何拉曼光譜儀。該軟體優化信號識別，並將拉曼光譜轉換為資料庫，促進分子指紋條碼的生成，以實現高效資料處理和大規模分析。CSDP已成功應用於農業和生物醫學測試，包括農藥、病毒、藥物、細菌、癌細胞和新冠病毒，能夠快速、便攜且強大地篩查各種應用，包括醫療場所的即時檢測（POCT）。

Medial temporal lobe epilepsy (MTLE) is a common type of focal epilepsy in adults, characterized by hippocampal sclerosis as the most prevalent pathology. Surgical resection for epilepsy can achieve seizure freedom in 60-80% of cases, leading to improved quality of life and restoration of normal social functioning in patients. The success of surgery relies on accurate preoperative localization and lateralization of the epileptic focus. Preoperative evaluation typically includes magnetic resonance imaging (MRI), prolonged video electroencephalogram (EEG), neuropsychological assessment, and positron emission tomography (PET). In patients with less evident hippocampal sclerosis, MRI findings may be subtle or atypical, posing challenges in determining the epileptic focus. FDG-PET imaging demonstrates metabolic decreases in 85-90% of temporal lobe epilepsy cases, providing valuable information for lateralizing the epileptic focus. Currently, FDG-PET is used to assist in lateralizing unilateral temporal lobe epilepsy, often relying on subjective comparison of asymmetry in PET signal intensity between the two sides of the medial temporal lobe. However, lacking standardized quantitative methods, visual assessment may not yield reliable results, particularly in cases with minimal asymmetry. This inadequacy in preoperative evaluation may necessitate invasive intracranial EEG implantation for further localization. Additionally, software accompanying PET scanners, based on normative databases within specific age ranges, is commonly utilized for reference. While quantitative, this approach may suffer from inaccuracies in brain region segmentation. Despite its sensitivity to glucose metabolism, FDG-PET suffers from poor spatial resolution and often requires fusion with MRI to delineate specific brain regions accurately. We propose a method that involves automated segmentation of high-quality MRI and fusion with FDG-PET images. Standardized uptake values (SUVs) of regions of interest (ROIs) are calculated using the mean gray matter FDG uptake as a reference for individual patients. We validated this approach using preoperative MRI and FDG-PET data from 95 patients with unilateral hippocampal sclerosis who underwent surgery. Machine learning classification based on ROI SUVs was used to verify the accuracy of epilepsy lateralization and compared with visual assessment. The method achieved 100% accuracy in lateralization in an external test set.

The intelligent induction building block system will transform the rehabilitation projects that cannot be achieved clinically with traditional building blocks by optimization and upgrade of traditional hand function rehabilitation tools, introduce hardware and software, and utilize Internet.

The intellegent system is integrated with three type of training, conical cup, insert board, and building blocks, and six modes of assessment, shoulders, elbows, wrists, hand fine movements, hand-eye coordination, and muscle endurance.

Among them, the training will combine occupational therapy theory, the interactive entertainment of mixed reality, and the gamification factors of game design theory to create a simulation game type and an action game type training course.

The training results data will be analyze through AI system and graphical data will be generated. Users can enter the service platform from different devices in a cross-platform manner through responsive web pages to understand the training results and usage status.

This technology optimization and upgrades of traditional rehabilitation building blocks, adds corresponding detection modules (including physical sensors, optical sensors, and hybrid sensors with the first two functions), and obtains hand signals through MR devices.

The image data and motion trajectory information of the hand and arm, plus signals from the Inertial Measurement Unit placed on the shoulder joint and auxiliary tools or devices, will be analyzed by algorithms to obtain more accurate assessment results.

AI data analysis and visual data presentation will be done by having AI analysis professionals in the team.

Visual data will help users quickly grasp the performance of the course, predict future progress, and help optimize reconstruction plans.

The products produced in the future will increase the willingness of rehabilitation patients to use and thewillingness of medical institutions to purchase.

在軟骨組織中，軟骨細胞 (Chondrocyte) 的比例約占整體組織的10 (w/v) %，由於軟骨組織缺乏血管及神經的分佈，導致軟骨細胞難以遷移到受損區域並且增生，所以軟骨受到破壞後，體內自我修復與再生的能力非常有限。因此，關節軟骨的修復，是一個目前急迫需要面對的問題。目前治療受損的軟骨，主要所採用的治療方式為馬賽克鑲嵌術(Mosaicplasty) 治療法，它是利用自體移植 (Autograft) 的方式，將其他部位的軟骨組織取下，來填補受損的區域。然而，此治療方法，因取其他部分的組織作為替代物，除了造成新的傷害外，還會有機械強度、數量、形狀大小的限制。為了解決上述Mosaicplasty 治療法的缺點，再生工程法是近年來大力發展的技術。
我們利用簡易交聯方法可以製作出方便價廉的3D支架，可以在體外生長軟骨後再移植回體內。
解決目前材料的機械強度不足問題，可以大量生產以及客製化製作各種形狀及大小。

"居家復健參與度低，因重複性活動、孤獨感和缺乏即時評估。
我們提出互動式 AI 教練系統，將表面肌電圖（sEMG）與遊戲化界面結合混合實境（MR）。
患者將在虛擬平台上與AI 教練、復健夥伴或治療師互動。接收視覺提示並完成動作後，根據肱橈肌、肱二頭肌和三角肌的電極收集 sEMG 信號生成音樂反饋。信號經神經網路工具解碼，提供肌肉活動資訊，識別上肢動作。初步結果顯示，SVM 及 CNN 模型能有效解碼上肢活動，受試者對音樂回饋表現正面，表明系統能提高患者的動力和參與度。系統旨在即時反饋和沉浸式環境中提升遠程醫療品質及可及性。未來研究將進行大規模臨床試驗，結合問卷調查和神經反饋改善使用者體驗及系統多樣性。"

"
Home-based rehabilitation can reduce travel costs and time, yet suffers from low participation due to repetitive activities, loneliness, and isolation (Nguyen et al., 2021). Additionally, the lack of real-time objective metrics to evaluate performance can decrease daily activity effectiveness. Addressing this, we propose an interactive rehabilitation system using surface electromyography (sEMG) and large language models (LLM) within a game-like mixed reality (MR) interface. This system creates virtual platform for real-time patient interactions with AI coach, partners or therapists. In the testing phase, users perform four poses representing pleasant chord structures, with visual prompts guiding them. Once completed, the chosen chords prompt the LLM to generate musical melodies as feedback. sEMG signals from electrodes on the arm provide crucial neuromuscular information (Reaz et al., 2006). The analyzed results show the Machine Learning (ML) model's feasibility in decoding muscular activity and an enhancement in user enjoyment with music feedback compared to conventional methods, suggesting heightened patient motivation. Our interactive system decodes real-time sEMG signals, generates melodies via prompted-LLM, and offers an immersive VR environment, enhancing telehealth quality and accessibility. Future studies will refine the system and conduct larger-scale clinical trials to improve versatility and user experience."

Hemophilia is a bleeding disorder caused by a deficiency in coagulation factors VIII and IX. Hemophilia A and B are the most common forms, affecting primarily males. Hemophilia A is the most prevalent, accounting for 80-85% of all cases. The World Federation of Hemophilia (WFH) reported in 2020 that globally there are 815,100 individuals with hemophilia, but only 347,026 are diagnosed, with 276,900 having severe hemophilia. Hemophilic arthropathy (HA) is a common complication, affecting over 90% of individuals with severe hemophilia due to recurrent joint bleeding (hemarthrosis). The challenges in diagnosing HA include:

(1) Delayed Diagnosis: Limited access to radiologists, especially in rural areas, leads to delays in diagnosis. Early-stage HA, characterized by soft-tissue swelling and subtle joint changes, is often missed in initial radiographic evaluations.
(2) Inconsistent Accuracy: Approximately 23% of ankle fractures are overlooked during the first radiographic evaluation due to factors such as insufficient training, outdated imaging technology, and anatomical variations.
(3) Resource Limitations: Advanced imaging techniques like MRI and CT scans, although more accurate, are not accessible in resource-limited settings due to high costs and infrastructure requirements.
(4) Need for Early Detection: Early detection of HA is crucial for initiating timely treatment to prevent irreversible joint damage and improve patient outcomes. Current methods often fail to identify early-stage HA effectively.
(5) Labor-Intensive Process: Manual interpretation of X-ray images by radiologists is time-consuming and subject to human error, affecting the efficiency and accuracy of diagnosis.

We use deep learning and machine learning approaches to SUCCESSFULLY develop an AI-enhanced diagnostic tool for automatic and accurate diagnosis of hemophilic arthropathy.

(1) Deep Learning (Inception V3): Extracts high-level features from X-ray images, eliminating the need for manual feature engineering.
(2) Machine Learning (SVM, KNN, XGBoost): Classifies X-ray images into 'Healthy' or 'Ill' categories. SVM achieved 97.78% accuracy and 0.99 AUC.

近年來，新興超級病毒和高抗藥性問題日益嚴重，促使尋找新一代抗生物質。病原體對多種抗生素耐藥性一直是嚴重問題。天然抗菌肽（AMPs）是一種功能性胜肽，相較於傳統藥物，不易引起細菌抗藥性，並具有抗癌和抗其他病原體的潛力。一般來說，新藥從開發到上市需花費4.6億美元，開發時間平均為10-15年，其中前期需3-6年，佔整體開銷約9成。本研究希望透過胜肽序列生成網路模型，設計全新具抗菌能力之胜肽，再以人工智慧加以鑑別有效且低毒性之候選胜肽，來進行合成與實地驗證，加速全新抗菌肽的設計與開發流程，預計可以加速前期開發時間並降低所需的費用。
本團隊收集梳理既有文獻與資料庫，以抗菌肽為例，整合後其所使用資料量為目前類似平台的兩到三倍，再利用深度學習卷積神經網路，建構蛋白質序列抗菌肽活性預測模型(mSystems, 2021)，並透過生成對抗網路模型（GAN)( IJMS, 2023），設計出與真實抗菌肽高度相似的序列，再以人工智慧技術篩選出高潛力低毒性之候選序列，進行後續的胜肽序列合成，精準化實地驗證的規模，目前已發現所合成的部分胜肽序列，具備抗癌症與抗黴菌(白色念珠菌)的能力同時亦維持著較低的溶血特性，後續正在與台大、台大藥學院與北醫等合作團隊進行更進一步的實驗驗證，目前已有四篇相關成果發表於與微生物與製藥相關之高影響力國際期刊，引用次數超過七十次。同時，因應不同功能性胜肽特性，評估使用集成學習（Ensemble AI）與資料增強（data augmentation）策略，來提升預測精準度；對於新興病毒如COVID-19等，則利用遷移式學習(Transfer Learning)方式，建立AMP/Anti-Viral Peptide (AVP)模型為基礎的功能鑑別模型，找出能對抗COVID-19及其他未來可能出現新興病毒的治療用胜肽。希望未來能將此一架構策略來延伸到不同治療用胜肽設計，加速其整體研發時程，提供我們更多針對未來新興疾病的可能對策，進而取代既有由化合物資料庫及大自然中，以海底撈針式曠日廢時的找尋方式。

本研究已發展出之技術成果簡述如下：
1. 完成AI治療用胜肽開發平台，包括
① 胜肽序列設計生成模型 ②集成式胜肽功能鑑別與 ③溶血預測平台(網站)
2. 進行全新抗藥性細菌抗菌肽之實地合成，及對抗特定病原及癌細胞之驗證 ，並已找出數條具抗菌、抗黴或抗癌潛力之低毒性胜肽。

「AI智能糖尿病足檢測系統」整合了AI外觀檢測、紅外線溫度檢測及智慧化感覺神經檢測三大核心技術。
- 相較傳統人工檢測，此系統擁有更高的準確性和效率，能更快速準確地偵測病患的潛在傷口和發炎問題。
- 系統設計自動化和智慧化，大幅降低醫療人力的投入，提升糖尿病足管理的可及性和頻率。

AcroCyte Therapeutics is a cell-based precision medicine company. Our current focus lies in utilizing our core technology, R3CE—rapid, reproducible, rare cell expansion—to assist clinical treatment and scientific research. R3CE is a robust cell proliferation and expansion platform that can enable 3D cell culture from any specimen within 2 weeks, even when the cells are extremely rare. We are a driven team of scientists, industry leaders, and entrepreneurs deeply committed to integrating 3D cell culture and medicine to drive changes in precision medicine and, in turn, save lives.