{"id":28060,"date":"2026-01-21T10:13:08","date_gmt":"2026-01-21T09:13:08","guid":{"rendered":"https:\/\/sano.science\/?p=28060"},"modified":"2026-01-21T10:25:37","modified_gmt":"2026-01-21T09:25:37","slug":"the-recording-of-mondays-sano-seminar-is-now-available","status":"publish","type":"post","link":"https:\/\/sano.science\/the-recording-of-mondays-sano-seminar-is-now-available\/","title":{"rendered":"The recording of Monday\u2019s Sano Seminar is now available\u00a0"},"content":{"rendered":"\n

The presentation focused on the development of trustworthy AI systems<\/strong> for early triage of acute ischemic stroke<\/strong> using EEG, with a particular emphasis on pre-hospital settings<\/strong> such as ambulances and on-scene emergency care. <\/p>\n\n\n\n

<\/div>\n\n\n\n

Key takeaways from the talk<\/strong> <\/p>\n\n\n

    \n
  • Project\u00a0goal:<\/strong>\u00a0to\u00a0develop\u00a0a\u00a0trustworthy\u00a0AI system for\u00a0early\u00a0ischemic\u00a0stroke\u00a0triage\u00a0based\u00a0on EEG,\u00a0suitable\u00a0for\u00a0use\u00a0in\u00a0pre-hospital\u00a0conditions.\u00a0<\/li>\n<\/ul>\n\n
      \n
    • Clinical\u00a0data<\/strong>\u00a0were\u00a0collected\u00a0within\u00a0the\u00a0ElectroStroke<\/strong>\u00a0and\u00a0AI\u00a0Stroke<\/strong>\u00a0clinical\u00a0trials\u00a0at\u00a0Amsterdam UMC<\/strong>,\u00a0using\u00a08-channel EEG with\u00a0dry\u00a0electrodes<\/strong>,\u00a0recorded\u00a0by\u00a0paramedics.\u00a0<\/li>\n<\/ul>\n\n
        \n
      • A major challenge was\u00a0EEG\u00a0signal\u00a0quality<\/strong>,\u00a0affected\u00a0by\u00a0motion\u00a0artifacts,\u00a0environmental\u00a0noise, and the\u00a0lack\u00a0of a\u00a0universal\u00a0processing\u00a0pipeline\u00a0for\u00a0low-density\u00a0EEG.\u00a0<\/li>\n<\/ul>\n\n
          \n
        • A\u00a0systematic\u00a0comparison\u00a0of EEG\u00a0artifact\u00a0removal\u00a0methods<\/strong>\u00a0was\u00a0performed,\u00a0including\u00a0ASR,\u00a0Wavelet\u00a0Transform, EMD<\/strong>, and\u00a0hybrid\u00a0approaches,\u00a0using\u00a0simulated, benchmark, and real\u00a0clinical\u00a0data.\u00a0<\/li>\n<\/ul>\n\n
            \n
          • Wavelet\u00a0Transform<\/strong>\u00a0performed\u00a0best\u00a0for\u00a0general\u00a0denoising\u00a0of\u00a0low-channel EEG,\u00a0while\u00a0EMD<\/strong>\u00a0better\u00a0preserved\u00a0pathological\u00a0patterns<\/strong>\u00a0such\u00a0as\u00a0sharp\u00a0waves\u00a0and\u00a0rhythmic\u00a0discharges.\u00a0<\/li>\n<\/ul>\n\n
              \n
            • Automatic\u00a0machine\u00a0learning (AutoML\u00a0\u2013\u00a0AutoPrognosis)<\/strong>\u00a0was\u00a0used\u00a0to\u00a0classify\u00a0ischemic\u00a0stroke\u00a0vs.\u00a0stroke\u00a0mimics<\/strong>,\u00a0enabling\u00a0fair and\u00a0reproducible\u00a0comparisons\u00a0between\u00a0processing\u00a0pipelines.\u00a0<\/li>\n<\/ul>\n\n
                \n
              • Although\u00a0standard performance\u00a0metrics\u00a0(accuracy, AUC)\u00a0were\u00a0similar\u00a0across\u00a0methods,\u00a0clinical\u00a0error\u00a0analysis<\/strong>\u00a0revealed\u00a0meaningful\u00a0differences\u00a0from a\u00a0patient’s\u00a0safety\u00a0perspective<\/strong>.\u00a0<\/li>\n<\/ul>\n\n
                  \n
                • In\u00a0pre-hospital\u00a0triage,\u00a0sensitivity\u00a0is\u00a0critical<\/strong>; a\u00a0hybrid\u00a0EMD +\u00a0Wavelet\u00a0approach<\/strong>\u00a0minimized\u00a0the\u00a0number\u00a0of\u00a0missed\u00a0strokes.\u00a0<\/li>\n<\/ul>\n\n
                    \n
                  • The system\u00a0incorporated\u00a0uncertainty\u00a0quantification<\/strong>\u00a0(both\u00a0aleatoric\u00a0and\u00a0epistemic) and a\u00a0rejection\u00a0mechanism<\/strong>\u00a0to\u00a0refer\u00a0uncertain\u00a0cases\u00a0to a\u00a0physician.\u00a0<\/li>\n<\/ul>\n\n
                      \n
                    • A\u00a0cascaded\u00a0model\u00a0architecture<\/strong>\u00a0was\u00a0proposed\u00a0to mirror\u00a0clinical\u00a0reasoning:\u00a0
                      Is\u00a0this\u00a0a\u00a0stroke? \u2192\u00a0Is\u00a0it\u00a0an\u00a0LVO? \u2192\u00a0Is\u00a0the\u00a0stroke\u00a0still\u00a0treatable?<\/em>,\u00a0while\u00a0accounting\u00a0for the\u00a0temporal\u00a0dynamics of EEG<\/strong>.\u00a0<\/li>\n<\/ul>\n\n\n
                      <\/div>\n\n\n\n

                      Watch the full seminar recording on YouTube:<\/strong> <\/p>\n\n\n\n

                      \n