[{"data":1,"prerenderedAt":236},["ShallowReactive",2],{"storyblokStory:en/insights/news/2026/imec-at-holst-centre-develops-chip-for-tenfold-neural-data-compression/?lang=en&version=published":3},{"story":4,"resolvedLinks":235},{"name":5,"created_at":6,"published_at":7,"updated_at":8,"id":9,"uuid":10,"content":11,"slug":226,"full_slug":227,"sort_by_date":14,"position":228,"tag_list":229,"is_startpage":29,"parent_id":230,"meta_data":14,"group_id":231,"first_published_at":232,"release_id":14,"lang":233,"path":14,"alternates":234,"default_full_slug":14,"translated_slugs":14},"Imec at Holst Centre develops chip for tenfold neural data compression","2026-07-06T11:32:37.811Z","2026-07-06T13:06:02.288Z","2026-07-06T13:06:02.305Z",195166035228990,"e9615811-ff9e-4d9f-8655-40149a5946cb",{"_uid":12,"ogImage":13,"ogTitle":15,"sections":18,"component":215,"metaTitle":15,"metaRobots":216,"ogSitename":15,"tagsDomain":217,"searchTitle":15,"metaKeywords":15,"overviewText":15,"metaCanonical":218,"ogDescription":15,"overviewImage":221,"overviewTitle":5,"metaDescription":15,"overviewCaption":15,"overviewEndDate":223,"overviewStartDate":224,"searchDescription":15,"overviewExternalLink":225},"d6dae121-b2c4-497c-9b72-8fe5b74fe029",{"id":14,"alt":14,"name":15,"focus":14,"title":14,"source":14,"filename":15,"copyright":14,"fieldtype":16,"meta_data":17},null,"","asset",{},[19],{"_uid":20,"component":21,"components":22,"containerWidth":212,"containerSpacingMarginX":213,"containerSpacingMarginY":214},"ae3fbaa5-b5a4-45c6-b81d-d157c9c75dc4","section-content",[23,36,59,74,116,176,186,201],{"_uid":24,"image":25,"title":5,"anchor":15,"titleTag":30,"component":31,"titleColor":32,"aspectRatio":33,"componentJustify":15,"contentSpacingPaddingX":34,"contentSpacingPaddingY":35},"27228045-5a9c-47c5-9089-9dd44b1be1ff",{"id":26,"alt":15,"name":15,"focus":15,"title":15,"source":15,"filename":27,"copyright":15,"fieldtype":16,"meta_data":28,"is_external_url":29},195166162663867,"https://a.storyblok.com/f/289273820865604/1280x853/801e5abd3a/_vbw.webp",{},false,"h2","content-image","blue","none","px-4 px-md-0","py-8 py-sm-16",{"_uid":37,"text":38,"anchor":15,"formId":15,"component":56,"linkStyle":32,"textAlign":52,"background":33,"titleColor":15,"componentWidth":57,"componentJustify":58,"contentSpacingPaddingX":34,"contentSpacingPaddingY":35},"59abc957-cf55-4229-a6d6-23cd6009bf3b",{"type":39,"content":40},"doc",[41,49],{"type":42,"attrs":43,"content":45},"heading",{"level":44,"textAlign":14},2,[46],{"text":47,"type":48},"The context: ","text",{"type":50,"attrs":51,"content":53},"paragraph",{"textAlign":52},"left",[54],{"text":55,"type":48},"Brain-computer interfaces (BCIs) could one day help restore movement, speech, or vision for people with paralysis or neurological disorders. But as these devices record from more and more neurons, they generate massive amounts of data, more than current wireless links can handle. Cramming in more bandwidth also means more heat, which risks damaging delicate brain tissue.","content-text","lg","center",{"_uid":60,"text":61,"image":65,"width":69,"anchor":15,"overlay":29,"titleTag":15,"component":70,"linkStyle":32,"textAlign":52,"titleColor":15,"aspectRatio":71,"captionColor":15,"contentWidth":72,"contentJustify":73,"contentBackgroundColor":33,"contentSpacingPaddingX":34,"contentSpacingPaddingY":35},"6b4618fe-236b-4316-a7d7-4c40113bde7c",{"type":39,"content":62},[63],{"type":50,"attrs":64},{"textAlign":14},{"id":66,"alt":15,"name":15,"focus":15,"title":15,"source":15,"filename":67,"copyright":15,"fieldtype":16,"meta_data":68,"is_external_url":29},195166546794951,"https://a.storyblok.com/f/289273820865604/9785x5504/44d42ef657/_vbw-0226-5059m2.jpg",{},"full","content-banner-image","16:9","md","start",{"_uid":75,"text":76,"anchor":15,"formId":15,"component":56,"linkStyle":32,"textAlign":52,"background":33,"titleColor":15,"componentWidth":57,"componentJustify":58,"contentSpacingPaddingX":34,"contentSpacingPaddingY":35},"3b826a86-9676-4a16-8dd6-9242815063da",{"type":39,"content":77},[78,90],{"type":50,"attrs":79,"content":80},{"textAlign":14},[81,83,88],{"text":82,"type":48},"Imec at Holst Centre developed a ",{"text":84,"type":48,"marks":85},"neuromorphic compressive telemetry (NCT) chip",[86],{"type":87},"bold",{"text":89,"type":48}," that solves this with two smart tricks:",{"type":91,"content":92},"bullet_list",[93,105],{"type":94,"content":95},"list_item",[96],{"type":50,"attrs":97,"content":98},{"textAlign":14},[99,103],{"text":100,"type":48,"marks":101},"Send-on-delta encoding: ",[102],{"type":87},{"text":104,"type":48},"Instead of constantly sampling brain signals (like a camera filming at a fixed frame rate), the chip only records when something actually changes. Since neurons fire rarely, this cuts out huge amounts of redundant data.",{"type":94,"content":106},[107],{"type":50,"attrs":108,"content":109},{"textAlign":14},[110,114],{"text":111,"type":48,"marks":112},"A smarter packaging system (eSER)",[113],{"type":87},{"text":115,"type":48},": Nearby brain signals often change together, so the chip bundles related events into one compact packet instead of sending each separately, avoiding delays and wasted bandwidth.",{"_uid":117,"text":118,"anchor":15,"formId":15,"component":56,"linkStyle":32,"textAlign":52,"background":33,"componentWidth":57,"componentJustify":58,"contentSpacingPaddingX":34,"contentSpacingPaddingY":35},"97429a4c-8a7c-4697-8baf-dd3e4f8dc558",{"type":39,"content":119},[120,125,130],{"type":42,"attrs":121,"content":122},{"level":44,"textAlign":14},[123],{"text":124,"type":48},"The results",{"type":50,"attrs":126,"content":127},{"textAlign":14},[128],{"text":129,"type":48},"Tested on real neural recordings across 384 channels, the chip:",{"type":91,"content":131},[132,143,154,165],{"type":94,"content":133},[134],{"type":50,"attrs":135,"content":136},{"textAlign":14},[137,141],{"text":138,"type":48,"marks":139},"Cut data volume by more than 10x",[140],{"type":87},{"text":142,"type":48},", even accounting for overhead",{"type":94,"content":144},[145],{"type":50,"attrs":146,"content":147},{"textAlign":14},[148,152],{"text":149,"type":48,"marks":150},"Kept signal quality high",[151],{"type":87},{"text":153,"type":48}," — even very faint spikes were reconstructed accurately",{"type":94,"content":155},[156],{"type":50,"attrs":157,"content":158},{"textAlign":14},[159,163],{"text":160,"type":48,"marks":161},"Used just 0.1 microwatts per channel",[162],{"type":87},{"text":164,"type":48}," — extremely low power, which matters for implants",{"type":94,"content":166},[167],{"type":50,"attrs":168,"content":169},{"textAlign":14},[170,174],{"text":171,"type":48,"marks":172},"Delivered ultra-precise timing",[173],{"type":87},{"text":175,"type":48}," (under 10 microseconds of variation), essential for tracking how neurons fire in relation to one another",{"_uid":177,"text":178,"image":182,"width":69,"anchor":15,"overlay":29,"component":70,"linkStyle":32,"textAlign":52,"titleColor":15,"aspectRatio":71,"captionColor":15,"contentWidth":72,"contentJustify":73,"contentBackgroundColor":33,"contentSpacingPaddingX":34,"contentSpacingPaddingY":35},"d6c6caf7-5a3c-4dca-9765-ff8c5df85095",{"type":39,"content":179},[180],{"type":50,"attrs":181},{"textAlign":14},{"id":183,"alt":15,"name":15,"focus":15,"title":15,"source":15,"filename":184,"copyright":15,"fieldtype":16,"meta_data":185,"is_external_url":29},195167238846995,"https://a.storyblok.com/f/289273820865604/8256x5504/bdcf2bfc2f/_vbw-0226-5031m1.jpg",{},{"_uid":187,"text":188,"anchor":15,"formId":15,"titleTag":15,"component":56,"linkStyle":32,"textAlign":52,"background":33,"titleColor":15,"captionColor":15,"componentWidth":57,"componentJustify":58,"contentSpacingPaddingX":34,"contentSpacingPaddingY":35},"d704374e-f156-4b6b-b3e4-9de198e6394e",{"type":39,"content":189},[190],{"type":50,"attrs":191,"content":192},{"textAlign":14},[193,195,199],{"text":194,"type":48},"The team is now adding an AI layer that automatically identifies the ~1% of neural signals that matter most, aiming to push compression up to 100x and pave the way for BCIs with 10,000 recording channels. This work was funded by an ERC Consolidator Grant led by ",{"text":196,"type":48,"marks":197},"Yao-Hong Liu",[198],{"type":87},{"text":200,"type":48},", Scientific Director at imec, and marks a concrete step toward BCIs that are both clinically viable and scalable.",{"src":202,"_uid":203,"loop":29,"image":204,"muted":29,"title":208,"anchor":15,"dialog":29,"autoplay":29,"controls":29,"titleTag":30,"component":209,"titleColor":32,"description":210,"titleBorder":211,"contentWidth":72,"contentJustify":58,"contentSpacingPaddingX":34,"contentSpacingPaddingY":35},"https://vimeo.com/1205420024","218e5f7f-e13a-4f98-8666-173c7775813f",{"id":205,"alt":15,"name":15,"focus":15,"title":15,"source":15,"filename":206,"copyright":15,"fieldtype":16,"meta_data":207,"is_external_url":29},195168636685020,"https://a.storyblok.com/f/289273820865604/8256x5504/2f79248389/_vbw-0226-5072-1.jpg",{},"Explanatory video: ","content-video","Intranet of Neurons- ERC- Yao-Hong Liu",true,"contain","mx-0","my-4 my-sm-8 my-lg-12","page-content",[],[],{"id":15,"url":15,"linktype":219,"fieldtype":220,"cached_url":15},"story","multilink",{"id":205,"alt":15,"name":15,"focus":15,"title":15,"source":15,"filename":206,"copyright":15,"fieldtype":16,"meta_data":222,"is_external_url":29},{},"2200-01-01 00:00","2026-07-13 00:00",{"id":15,"url":15,"linktype":219,"fieldtype":220,"cached_url":15},"imec-at-holst-centre-develops-chip-for-tenfold-neural-data-compression","en/insights/news/2026/imec-at-holst-centre-develops-chip-for-tenfold-neural-data-compression",0,[],172906887887506,"ea6af44a-0860-44f9-bbc5-bce868848226","2026-07-06T11:45:45.572Z","default",[],[],1783343324397]