[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-84343-en":3,"doc-seo-84343-105":30,"detail-sidebar-cat-0-en-105":91},{"code":4,"msg":5,"data":6},0,"success",{"doc_id":7,"user_id":8,"nickname":9,"user_avatar":10,"doc_module":4,"category_id":11,"category_name":12,"doc_title":13,"doc_description":14,"doc_content":15,"file_id":16,"file_url":17,"file_type":18,"file_size":19,"view_count":20,"is_deleted":4,"is_public":21,"is_downloadable":21,"audit_status":21,"page_count":22,"language":23,"language_code":24,"site_id":25,"html_lang":24,"table_of_contents":26,"faqs":27,"seo_title":13,"seo_description":14,"update_tm":28,"read_time":29},84343,1099514068365,"Aurelia","https://ap-avatar.wpscdn.com/avatar/10000253d8d9f28188e?_k=1776742907772140068",8,"Research & Report","Progression as Latent Drift: Generative Forecasting of Slow-Evolving Pathologies","Forecasting the future anatomy of slow-evolving neurodegenerative diseases supports earlier, targeted intervention and can strengthen clinical trial design, yet longitudinal MRI progression signals are extremely subtle. In this low-signal regime, direct transfer of modern generative sequence models fails because optimization is dominated by stable baseline anatomy and confounded by dense, sample-specific nuisance variation. The work analyzes two failure modes—identity collapse and continuous interpolation trap—and introduces Latent Drift to learn progression as compressed latent change with finite scalar quantization to suppress nuisance fluctuations while preserving structural drift.","arXiv :2607 .08270v 1 [ cs .CV] 9 Jul 2026  \nProgression as Latent Drift: Generative Forecasting of Slow-Evolving Pathologies  \nYuxiang Feng 1 ,3⋆‡, Juncheng Wang2⋆, Chao Xu3 ,4 , Wenlong Hou2 Huihan Wang2 , Yijie Qian 1 ,3 , Yang Liu3 ,4 , Baigui Sun3 ,4 Yong Liu 1†, and Shujun Wang2†  \n1 Zhejiang University, Hangzhou, China  \n2 The Hong Kong Polytechnic University, Hong Kong, China  \n3 IROOTECH TECHNOLOGY, China  \n4 Wolf 1069 b Lab, Sany Group, China ⋆ Equal contribution. † Corresponding authors.  \n{[fengyx@zju.edu.cn](fengyx@zju.edu.cn), [wjc2830@gmail.com}](wjc2830@gmail.com})⋆ , {[yongliu@iipc.zju.edu.cn](yongliu@iipc.zju.edu.cn),  \n[shu-jun.wang@polyu.edu.hk}](shu-jun.wang@polyu.edu.hk})†  \nAbstract. Forecasting the future anatomy of slow-evolving neurodegenerative diseases could enable earlier, more targeted intervention and improve clinical trial design, but it remains challenging because true progression signals are subtle in longitudinal MRI. In this low-signal regime, transferring modern generative sequence models directly is unreliable:  \ntraining is dominated by stable baseline anatomy and confounded by dense, sample-specific nuisance variation. We first provide a theoretical analysis that explains these failures through two modes. Identity collapse occurs when optimization is driven toward reproducing the current anatomy, which prevents the model from learning faint temporal change.  \nThe continuous interpolation trap arises when standard smooth networks cannot separate localized biological drift from pervasive noise, which leads to spurious changes that diffuse across the volume. To address both issues, we propose Latent Drift, a progressive generative framework that learns change in a compressed semantic representation rather than synthesizing full-resolution anatomy. This design removes pixel-level identity from the prediction target and concentrates model capacity on progression-relevant dynamics. We further apply Finite Scalar Quantization to the learned change representation, which suppresses small, highfrequency nuisance fluctuations while preserving consistent structural drift. Experiments on longitudinal 3D brain MRI show that Latent Drift improves patient-specific neuro-forecasting over diffusion and autoregressive transformer baselines across generative fidelity and clinically relevant evaluation metrics. Project page: [https://cutepkq.github.io/latent-drift](https://cutepkq.github.io/latent-drift).  \nKeywords: Slow-Evolving Pathologies · Generative Model · Brain Simulation  \n1 Introduction  \nDynamic simulation of organ evolution [7, 12, 50] is essential for forecasting disease trajectories, particularly for neurodegeneration and other slow-evolving  \n‡ This work was conducted in collaboration with IROOTECH TECHNOLOGY.  \n2 Y. Feng et al.  \nFig. 1: Progression as Latent Drift. (a) Task Formulation: Autoregressive forecasting of future MRI states in a compressed semantic latent space. (b) Direct Future Generation: Predicting the absolute future state induces Identity Collapse, as stationary background anatomy overwhelms microscopic biological changes. (c) Latent Drift Generation (Ours): Predicting the temporal residual (∆z) bypasses this collapse, isolating the true pathological trajectory. (d) FSQ as a Topological Dead-Zone: To break the Continuous Interpolation Trap caused by dense imaging noise, Finite Scalar Quantization (FSQ) serves as a non-Lipschitz filter that annihilates nuisance variations while preserving sparse semantic drift.  \nbrain pathologies that progress along largely irreversible courses. Because neuronal loss and downstream structural changes (e.g., cortical thinning and ventricular enlargement) cannot be recovered [1, 5], substantial tissue damage may already be present by the time symptoms become clinically apparent [4,13] . This delayed observability narrows the effective window for intervention and complicates the development and evaluation of disease-modifying therapies, which in","cbCairuOvZxKYxz6","https://ap.wps.com/l/cbCairuOvZxKYxz6","pdf",3821412,4,1,31,"English","en",105,"# Introduction\n## Problem: low-signal progression in longitudinal MRI\n## Generative forecasting for disease trajectories\n# Method Overview\n## Latent Drift framework\n## Finite Scalar Quantization (FSQ)\n# Failure Analysis\n## Identity collapse\n## Continuous interpolation trap","[{\"question\":\"Why is generative forecasting difficult for slow-evolving neurodegenerative diseases?\",\"answer\":\"Because current and future MRI states are statistically very similar, so disease progression signals are much weaker than nuisance variation and stable baseline anatomy.\"},{\"question\":\"What is identity collapse in the context of forecasting future MRI states?\",\"answer\":\"When predicting absolute future states, optimization tends to reproduce the current anatomy, preventing the model from learning faint temporal changes relevant to pathology progression.\"},{\"question\":\"How does Latent Drift address these failures?\",\"answer\":\"Latent Drift predicts temporal residual change in a compressed semantic latent representation rather than synthesizing full-resolution anatomy, and uses finite scalar quantization to suppress small high-frequency nuisance fluctuations while preserving structural drift.\"}]",1784194945,78,{"code":4,"msg":31,"data":32},"ok",{"site_id":25,"language":24,"slug":33,"title":13,"keywords":34,"description":14,"schema_data":35,"social_meta":86,"head_meta":88,"extra_data":90,"updated_unix":28},"progression-as-latent-drift-generative-forecasting-of-slow-evolving-pathologies","",{"@graph":36,"@context":85},[37,53,68],{"@type":38,"itemListElement":39},"BreadcrumbList",[40,44,48,51],{"item":41,"name":42,"@type":43,"position":21},"https://docshare.wps.com","Home","ListItem",{"item":45,"name":46,"@type":43,"position":47},"https://docshare.wps.com/document/","Document",2,{"item":49,"name":12,"@type":43,"position":50},"https://docshare.wps.com/document/research-report/",3,{"item":52,"name":13,"@type":43,"position":20},"https://docshare.wps.com/document/progression-as-latent-drift-generative-forecasting-of-slow-evolving-pathologies/84343/",{"url":52,"name":13,"@type":54,"author":55,"headline":13,"publisher":57,"fileFormat":60,"inLanguage":24,"description":14,"dateModified":61,"datePublished":62,"encodingFormat":60,"isAccessibleForFree":63,"interactionStatistic":64},"DigitalDocument",{"name":9,"@type":56},"Person",{"url":41,"name":58,"@type":59},"DocShare","Organization","application/pdf","2026-07-28","2026-07-16",true,{"@type":65,"interactionType":66,"userInteractionCount":20},"InteractionCounter",{"@type":67},"ViewAction",{"@type":69,"mainEntity":70},"FAQPage",[71,77,81],{"name":72,"@type":73,"acceptedAnswer":74},"Why is generative forecasting difficult for slow-evolving neurodegenerative diseases?","Question",{"text":75,"@type":76},"Because current and future MRI states are statistically very similar, so disease progression signals are much weaker than nuisance variation and stable baseline anatomy.","Answer",{"name":78,"@type":73,"acceptedAnswer":79},"What is identity collapse in the context of forecasting future MRI states?",{"text":80,"@type":76},"When predicting absolute future states, optimization tends to reproduce the current anatomy, preventing the model from learning faint temporal changes relevant to pathology progression.",{"name":82,"@type":73,"acceptedAnswer":83},"How does Latent Drift address these failures?",{"text":84,"@type":76},"Latent Drift predicts temporal residual change in a compressed semantic latent representation rather than synthesizing full-resolution anatomy, and uses finite scalar quantization to suppress small high-frequency nuisance fluctuations while preserving structural drift.","https://schema.org",{"og:url":52,"og:type":87,"og:title":13,"og:site_name":58,"og:description":14},"article",{"robots":89,"canonical":52},"index,follow",{"doc_id":7,"site_id":25},{"code":4,"msg":5,"data":92},[93,97,101,105,110,115,120,123,128,131,135],{"id":21,"doc_module":4,"doc_module_name":46,"category_name":94,"show_sort_weight":95,"slug":96},"Story & Novel",90,"story-novel",{"id":47,"doc_module":4,"doc_module_name":46,"category_name":98,"show_sort_weight":99,"slug":100},"Literature",80,"literature",{"id":20,"doc_module":4,"doc_module_name":46,"category_name":102,"show_sort_weight":103,"slug":104},"Exam",70,"exam",{"id":106,"doc_module":4,"doc_module_name":46,"category_name":107,"show_sort_weight":108,"slug":109},5,"Comic",60,"comic",{"id":111,"doc_module":4,"doc_module_name":46,"category_name":112,"show_sort_weight":113,"slug":114},6,"Technology",50,"technology",{"id":116,"doc_module":4,"doc_module_name":46,"category_name":117,"show_sort_weight":118,"slug":119},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":46,"category_name":12,"show_sort_weight":121,"slug":122},30,"research-report",{"id":124,"doc_module":4,"doc_module_name":46,"category_name":125,"show_sort_weight":126,"slug":127},9,"Religion & Spirituality",20,"religion-spirituality",{"id":126,"doc_module":4,"doc_module_name":46,"category_name":129,"show_sort_weight":126,"slug":130},"World Cup","world-cup",{"id":132,"doc_module":4,"doc_module_name":46,"category_name":133,"show_sort_weight":132,"slug":134},10,"Lifestyle","lifestyle",{"id":136,"doc_module":4,"doc_module_name":46,"category_name":137,"show_sort_weight":106,"slug":138},19,"General","general"]