[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-seo-242664-105":3,"detail-sidebar-cat-1-en-105":81,"doc-detail-242664-en":127},{"code":4,"msg":5,"data":6},0,"ok",{"site_id":7,"language":8,"slug":9,"title":10,"keywords":11,"description":12,"schema_data":13,"social_meta":74,"head_meta":76,"extra_data":78,"updated_unix":80},105,"en","ft-mdt-extracting-decision-trees-from-medical-texts-via-a-novel-low-rank-adaptation-method","FT-MDT - Extracting Decision Trees from Medical Texts via a Novel Low-rank Adaptation Method","","Knowledge of medical decision processes, modeled as medical decision trees (MDTs), is essential for clinical decision support systems. Existing MDT construction depends heavily on manual annotation, making it slow and labor-intensive. PI-LoRA (Path-Integrated LoRA) automatically extracts MDTs from clinical guidelines and textbooks by integrating gradient path information to capture synergistic effects across model modules. The method allocates ranks to critical components and prunes less important ones, improving efficiency and accuracy. Extensive experiments on medical guideline datasets show strong gains over prior parameter-efficient fine-tuning approaches for Text2MDT, achieving state-of-the-art results with reduced model complexity suitable for resource-limited clinical settings.",{"@graph":14,"@context":73},[15,34,56],{"@type":16,"itemListElement":17},"BreadcrumbList",[18,23,27,31],{"item":19,"name":20,"@type":21,"position":22},"https://docshare.wps.com","Home","ListItem",1,{"item":24,"name":25,"@type":21,"position":26},"https://docshare.wps.com/template/","Template",2,{"item":28,"name":29,"@type":21,"position":30},"https://docshare.wps.com/template/presentations/","Presentations",3,{"item":32,"name":10,"@type":21,"position":33},"https://docshare.wps.com/template/ft-mdt-extracting-decision-trees-from-medical-texts-via-a-novel-low-rank-adaptation-method/242664/",4,{"url":32,"name":10,"@type":35,"image":36,"author":41,"headline":10,"publisher":44,"fileFormat":47,"inLanguage":8,"description":12,"dateModified":48,"datePublished":49,"encodingFormat":47,"isAccessibleForFree":50,"interactionStatistic":51},"DigitalDocument",{"url":37,"@type":38,"width":39,"height":40},"https://docshare.wps.com/thumbnails/ft-mdt-extracting-decision-trees-from-medical-texts-via-a-novel-low-rank-adaptation-method/242664.png","ImageObject",442,249,{"name":42,"@type":43},"McGucket","Person",{"url":19,"name":45,"@type":46},"DocShare","Organization","application/pdf","2026-10-17","2026-09-12",true,{"@type":52,"interactionType":53,"userInteractionCount":55},"InteractionCounter",{"@type":54},"ViewAction",9,{"@type":57,"mainEntity":58},"FAQPage",[59,65,69],{"name":60,"@type":61,"acceptedAnswer":62},"What problem does PI-LoRA address in constructing medical decision trees?","Question",{"text":63,"@type":64},"It addresses the reliance of current MDT construction on slow and labor-intensive manual annotation by enabling automatic extraction from clinical guidelines and textbooks.","Answer",{"name":66,"@type":61,"acceptedAnswer":67},"How does PI-LoRA improve rank allocation compared with prior LoRA methods?",{"text":68,"@type":64},"It integrates gradient path information to capture synergistic effects between modules, ensuring more effective and reliable rank allocation by giving appropriate ranks to critical modules and pruning less important ones.",{"name":70,"@type":61,"acceptedAnswer":71},"What evidence is provided that PI-LoRA performs better for the Text2MDT task?",{"text":72,"@type":64},"Experiments on medical guideline datasets show that PI-LoRA outperforms existing parameter-efficient fine-tuning approaches, delivering better accuracy with substantially reduced model complexity and lightweight architecture.","https://schema.org",{"og:url":32,"og:type":75,"og:title":10,"og:site_name":45,"og:description":12},"article",{"robots":77,"canonical":32},"index,follow",{"doc_id":79,"site_id":7},242664,1791726976,{"code":4,"msg":82,"data":83},"success",[84,88,93,98,103,108,113,118,123],{"id":85,"doc_module":22,"doc_module_name":25,"category_name":29,"show_sort_weight":86,"slug":87},11,90,"presentations",{"id":89,"doc_module":22,"doc_module_name":25,"category_name":90,"show_sort_weight":91,"slug":92},12,"Resumes",80,"resumes",{"id":94,"doc_module":22,"doc_module_name":25,"category_name":95,"show_sort_weight":96,"slug":97},14,"Invoices",70,"invoices",{"id":99,"doc_module":22,"doc_module_name":25,"category_name":100,"show_sort_weight":101,"slug":102},15,"Posters",60,"posters",{"id":104,"doc_module":22,"doc_module_name":25,"category_name":105,"show_sort_weight":106,"slug":107},16,"Social Media",50,"social-media",{"id":109,"doc_module":22,"doc_module_name":25,"category_name":110,"show_sort_weight":111,"slug":112},17,"Forms",40,"forms",{"id":114,"doc_module":22,"doc_module_name":25,"category_name":115,"show_sort_weight":116,"slug":117},18,"Letters",30,"letters",{"id":119,"doc_module":22,"doc_module_name":25,"category_name":120,"show_sort_weight":121,"slug":122},21,"Paper Templates",5,"papers-templates",{"id":124,"doc_module":22,"doc_module_name":25,"category_name":125,"show_sort_weight":4,"slug":126},158,"General","general-158",{"code":4,"msg":82,"data":128},{"doc_id":79,"user_id":129,"nickname":42,"user_avatar":130,"doc_module":22,"category_id":85,"category_name":29,"doc_title":10,"doc_description":12,"doc_content":131,"file_id":132,"file_url":133,"file_type":134,"file_size":135,"view_count":55,"is_deleted":4,"is_public":22,"is_downloadable":22,"audit_status":22,"page_count":89,"language":136,"language_code":8,"site_id":7,"html_lang":8,"table_of_contents":137,"faqs":138,"seo_title":139,"seo_description":12,"update_tm":140,"read_time":33},1236954412713,"https://us-avatar.wpscdn.com/davatar_29158cc5080c5b710cf443261637dec0","FT-MDT: Extracting Decision Trees from Medical Texts via a Novel  \nLow-rank Adaptation Method  \nYuheng Li1,∗, Jiechao Gao2,†Wei Han3 , Wenwen Ouyang4 , Wei Zhu5,†, Hui Yi Leong6  \n1Johns Hopkins University, Baltimore, MD, United States 2 Stanford University, Stanford, CA, United States  \n3Independent Researcher, Austin, TX, United States. Email: [palebluedot.milkyway@gmail.com](palebluedot.milkyway@gmail.com)  \n4Carnegie Mellon University, Pittsburgh, PA, United States 5University of Hong Kong, Hong Kong, HK, China  \n6University of Chicage, Chicago, IL, United States  \nAbstract  \nKnowledge of the medical decision process, which can be modeled as medical decision trees (MDTs), is critical to building clinical decision support systems. However, current MDT construction methods rely heavily on time-consuming and laborious manual annotation. To address this challenge, we propose PI-LoRA (Path-Integrated LoRA), a novel lowrank adaptation method for automatically extracting MDTs from clinical guidelines and textbooks. We integrate gradient path information to capture synergistic effects between different modules, enabling more effective and reliable rank allocation. This framework ensures that the most critical modules receive appropriate rank allocations while less important ones are pruned, resulting in a more efficient and accurate model for extracting medical decision trees from clinical texts. Extensive experiments on medical guideline datasets demonstrate that our PI-LoRA method significantly outperforms existing parameter-efficient fine-tuning approaches for the Text2MDT task, achieving better accuracy with substantially reduced model complexity. The proposed method achieves state-of-the-art results while maintaining a lightweight architecture, making it particularly suitable for clinical decision support systems where computational resources may be limited.  \n1 Introduction  \nMedical decision processes are critical to building effective clinical decision support systems, and these processes can be effectively modeled as medical decision trees (MDTs) . The ability to automatically extract MDTs from clinical guidelines and textbooks would significantly reduce the reliance on time-consuming manual annotation while enabling the development of more robust decision  \n∗ These authors contribute equally.  \n† Corresponding author. For any inquiries, please contact: [michaelwzhu91@gmail.com](michaelwzhu91@gmail.com), [jiechao@stanford.edu](jiechao@stanford.edu).  \nsupport tools. However, the complex hierarchical nature of medical decision-making presents unique challenges for automated extraction, requiring sophisticated natural language processing techniques that can capture both the structural and semantic aspects of clinical guidelines.  \nPrevious research (Zhu et al., 2024a) on the Text2MDT task has primarily focused on either end-to-end frameworks using large language models or pipeline approaches, without fully investigating parameter-efficient fine-tuning (PEFT) methods (Ding et al., 2022) for this specific application. In addition, while PEFT methods like LoRA (Hu et al., 2021) have shown promise in reducing model complexity, existing LoRA-based approaches for Text2MDT suffer from critical limitations in rank allocation. Methods such as AdaLoRA (Zhang et al., 2023a) rely on sensitivity-based importance scores that are unreliable as they only consider how a single parameter change affects the model under the assumption that no other parameters change. Similarly, approaches like SoRA (Ding et al., 2023) and SaLoRA (Hu et al., 2023) use architectural parameters that cannot reliably reflect the quality or importance of LoRA ranks, leading to suboptimal performance in the context of medical decision tree extraction where precise structural representation is essential.  \nTo address these limitations, we propose PILoRA (Path-Integrated LoRA), a novel low-rank adaptation method that overcomes the deficiencies of existing rank allocati","cbCainFVoqp5Hodt","https://ap.wps.com/l/cbCainFVoqp5Hodt","pdf",277211,"English","# Introduction\n## Related works\n### Tree data extraction from text","[{\"question\":\"What problem does PI-LoRA address in constructing medical decision trees?\",\"answer\":\"It addresses the reliance of current MDT construction on slow and labor-intensive manual annotation by enabling automatic extraction from clinical guidelines and textbooks.\"},{\"question\":\"How does PI-LoRA improve rank allocation compared with prior LoRA methods?\",\"answer\":\"It integrates gradient path information to capture synergistic effects between modules, ensuring more effective and reliable rank allocation by giving appropriate ranks to critical modules and pruning less important ones.\"},{\"question\":\"What evidence is provided that PI-LoRA performs better for the Text2MDT task?\",\"answer\":\"Experiments on medical guideline datasets show that PI-LoRA outperforms existing parameter-efficient fine-tuning approaches, delivering better accuracy with substantially reduced model complexity and lightweight architecture.\"}]","FT-MDT - Extracting Decision Trees from Medical Texts via a Novel Low-rank Adaptation Method | PDF",1789192983]