دوره 22، شماره 2 - ( 9-1404 )
جلد 22 شماره 2 صفحات 47-42 |
برگشت به فهرست نسخه ها
Download citation:
BibTeX | RIS | EndNote | Medlars | ProCite | Reference Manager | RefWorks
Send citation to:
BibTeX | RIS | EndNote | Medlars | ProCite | Reference Manager | RefWorks
Send citation to:
Fateh H, Asgharpour M, Khayatajami M, Jameie S B, Derakhshesh A, Farhadi M, et al . AI-Driven Optimization of Long-Term Potentiation Protocols in an Experimental Model of Alzheimer’s Disease. ASWTR 2025; 22 (2) :42-47
URL: http://icml.ir/article-1-681-fa.html
URL: http://icml.ir/article-1-681-fa.html
فاتح هاله، اصغرپور مانیا، خیاط عجمی مجتبی، جامعی سید بهنام الدین، درخشش آریا، فرهادی مونا، و همکاران.. AI-Driven Optimization of Long-Term Potentiation Protocols in an Experimental Model of Alzheimer’s Disease. Advances in Skin, Wound and Tissue Repair. 1404; 22 (2) :42-47
هاله فاتح 
، مانیا اصغرپور ، مجتبی خیاط عجمی ، سید بهنام الدین جامعی ، آریا درخشش ، مونا فرهادی ، سبحان کاظمی ، حسام الدین علامه ، نسرین حسینی*
، مانیا اصغرپور ، مجتبی خیاط عجمی ، سید بهنام الدین جامعی ، آریا درخشش ، مونا فرهادی ، سبحان کاظمی ، حسام الدین علامه ، نسرین حسینی*
چکیده: (24 مشاهده)
Background: Alzheimer’s disease (AD), one of the most prevalent neurodegenerative disorders, is marked by progressive memory loss and cognitive decline. With the rapid advancement of artificial intelligence (AI), data-driven approaches have become increasingly important in both basic and clinical neuroscience research. While animal models remain indispensable for elucidating disease mechanisms and testing therapeutic strategies, such experiments are often costly and time-consuming. This study, therefore, sought to refine the experimental protocol for inducing long-term potentiation (LTP) in an animal model of AD through AI-driven data mining techniques.
Methods: Electrophysiological data were obtained from male Wistar rats (n = 36) divided into three groups: control, sham-operated, and AD. The AD model was induced via selective lesioning of the Basal Nucleus of Meynert (NBM) using ibotenic acid. Data preprocessing and analysis were performed in Python. Feature selection and ranking were conducted using Mutual Information, Gain Ratio, and Gini Index, followed by correlation-based filtering through the visualization of heatmaps. Features with low predictive power or high redundancy were systematically excluded to construct a more efficient classification framework.
Results: Feature importance analysis identified Mean Per.PSA.B6–B8 and Mean Per.Slop.B6–B8 as highly predictive variables, whereas pre-tetanic measures (Mean Per.PSA.B3–B4 and Mean Per.Slop.B3–B4) contributed minimally and were excluded. Likewise, high-intensity stimulation features (Mean PSA600–1000 and Mean Slop600–1000) demonstrated more substantial predictive value compared to low-intensity counterparts (<600 μA), supporting their retention in the optimized protocol. Correlation analysis confirmed these findings by highlighting redundancy among low-importance features.
Conclusions: The integration of feature importance metrics with correlation-based filtering enabled the identification of key electrophysiological markers while eliminating redundant variables from the LTP protocol. This optimization enhanced the accuracy, stability, and interpretability of machine learning models, while simultaneously reducing experimental costs, duration, and data collection requirements. Dimensionality reduction further improved computational efficiency and predictive performance, particularly in complex architectures.
Significance: This study introduces the first evidence-based, machine learning–guided protocol optimization framework in Alzheimer’s research, bridging computational intelligence with neurophysiological modeling to accelerate translational discovery.
Methods: Electrophysiological data were obtained from male Wistar rats (n = 36) divided into three groups: control, sham-operated, and AD. The AD model was induced via selective lesioning of the Basal Nucleus of Meynert (NBM) using ibotenic acid. Data preprocessing and analysis were performed in Python. Feature selection and ranking were conducted using Mutual Information, Gain Ratio, and Gini Index, followed by correlation-based filtering through the visualization of heatmaps. Features with low predictive power or high redundancy were systematically excluded to construct a more efficient classification framework.
Results: Feature importance analysis identified Mean Per.PSA.B6–B8 and Mean Per.Slop.B6–B8 as highly predictive variables, whereas pre-tetanic measures (Mean Per.PSA.B3–B4 and Mean Per.Slop.B3–B4) contributed minimally and were excluded. Likewise, high-intensity stimulation features (Mean PSA600–1000 and Mean Slop600–1000) demonstrated more substantial predictive value compared to low-intensity counterparts (<600 μA), supporting their retention in the optimized protocol. Correlation analysis confirmed these findings by highlighting redundancy among low-importance features.
Conclusions: The integration of feature importance metrics with correlation-based filtering enabled the identification of key electrophysiological markers while eliminating redundant variables from the LTP protocol. This optimization enhanced the accuracy, stability, and interpretability of machine learning models, while simultaneously reducing experimental costs, duration, and data collection requirements. Dimensionality reduction further improved computational efficiency and predictive performance, particularly in complex architectures.
Significance: This study introduces the first evidence-based, machine learning–guided protocol optimization framework in Alzheimer’s research, bridging computational intelligence with neurophysiological modeling to accelerate translational discovery.
ارسال پیام به نویسنده مسئول
| بازنشر اطلاعات | |
 |
این مقاله تحت شرایط Creative Commons Attribution-NonCommercial 4.0 International License قابل بازنشر است. |