BEGIN:VCALENDAR VERSION:2.0 PRODID:-//linuxsoftware.nz//NONSGML Joyous v1.4//EN BEGIN:VEVENT SUMMARY:Temporal Phenotyping of ALS Patients using Machine Learning DTSTART:20250704T120000Z DTEND:20250704T140000Z DTSTAMP:20260704T215518Z UID:6d7767b8-1aab-409e-8a9d-8b269c504b3d SEQUENCE:1 CREATED:20250702T075546Z DESCRIPTION: Password: 561232 Amyotrophic Lateral Sclerosis (ALS) is a fa tal neurodegenerative disease characterized by highly heterogeneous and ra pidly progressing motor decline. Despite extensive research\, predicting A LS progression remains a major clinical challenge\, limiting timely interv entions such as Non-Invasive Ventilation (NIV) and Percutaneous Endoscopic Gastrostomy (PEG). This thesis addresses this challenge by applying machi ne learning (ML)-based temporal phenotyping to characterize evolving patie nt profiles from static and longitudinal clinical data\, with the objectiv e of uncovering distinct progression trajectories and improving prognosis prediction.Specifically\, the T-Phenotype algorithm is adapted to the Lisb on ALS Clinic dataset\, leveraging phenotypic predictive clustering to gro up patients by both outcome and disease trajectory similarities. Comprehen sive preprocessing strategies were developed to handle irregular clinical event timing\, missing data\, variable-length time series\, and class imba lance. The algorithm identified clinically meaningful phenotypes associate d with the need for NIV (endpoint C1) and PEG (endpoint C3)\, with optimal performance in shorter observation windows (6–12 months)\, revealing di stinct risk profiles aligned with respiratory and bulbar function decline. Combining endpoints into a multiclass task (C1+C3) highlighted challenges with class imbalance and reduced prediction performance\, although still produced interpretable phenotypes. Temporal analysis of cluster transition s illustrated the model’s ability to capture diverse ALS progression pat terns dynamically\, emphasizing the disease'\;s heterogeneity and the limitations of static classifications. Although the T-Phenotype model had satisfactory results in predictive accuracy and clustering interpretabilit y for the binary endpoints (Hprc metric around 80%)\, its performance decl ined for the combined endpoint (Hprc metric around 65%)\, its performance declined for the combined endpoint. Additionally\, it lacked sensitivity t o clinical improvements post-intervention\, underscoring the need for rich er datasets and methodological rigor.The potential of ML-based temporal ph enotyping was demonstrated as a valuable tool for understanding ALS progre ssion and supporting personalized prognosis and intervention planning. By integrating temporal dynamics into patient stratification\, this approach advances data-driven ALS monitoring and highlights critical considerations for future research in this domain. LAST-MODIFIED:20250702T075546Z LOCATION:Online URL:http://df.vps.tecnico.ulisboa.pt/en/events/temporal-phenotyping-of-als -patients-using-machine-learning/ X-ALT-DESC;FMTTYPE=text/html:

Password: 561232 < /p>

Amyotrophic La teral Sclerosis (ALS) is a fatal neurodegenerative disease characterized b y highly heterogeneous and rapidly progressing motor decline. Despite exte nsive research\, predicting ALS progression remains a major clinical chall enge\, limiting timely interventions such as Non-Invasive Ventilation (NIV ) and Percutaneous Endoscopic Gastrostomy (PEG). This thesis addresses thi s challenge by applying machine learning (ML)-based temporal phenotyping t o characterize evolving patient profiles from static and longitudinal clin ical data\, with the objective of uncovering distinct progression trajecto ries and improving prognosis prediction.

Specifically\, the T-Phenotype algorithm is adapted to the Lisbon ALS Cli nic dataset\, leveraging phenotypic predictive clustering to group patient s by both outcome and disease trajectory similarities. Comprehensive prepr ocessing strategies were developed to handle irregular clinical event timi ng\, missing data\, variable-length time series\, and class imbalance. The algorithm identified clinically meaningful phenotypes associated with the need for NIV (endpoint C1) and PEG (endpoint C3)\, with optimal performan ce in shorter observation windows (6–12 months)\, revealing distinct ris k profiles aligned with respiratory and bulbar function decline.

Combining endpoints into a multiclass task (C1+C3) highlighted challenges with class imbalance and reduced prediction performance\, although still produced interpretable phenotypes. Temporal analysis of cluster transition s illustrated the model’s ability to capture diverse ALS progression pat terns dynamically\, emphasizing the disease'\;s heterogeneity and the limitations of static classifications. Although the T-Phenotype model had satisfactory results in predictive accuracy and clustering interpretabilit y for the binary endpoints (Hprc metric around 80%)\, its performance decl ined for the combined endpoint (Hprc metric around 65%)\, its performance declined for the combined endpoint. Additionally\, it lacked sensitivity t o clinical improvements post-intervention\, underscoring the need for rich er datasets and methodological rigor.

Th e potential of ML-based temporal phenotyping was demonstrated as a valuabl e tool for understanding ALS progression and supporting personalized progn osis and intervention planning. By integrating temporal dynamics into pati ent stratification\, this approach advances data-driven ALS monitoring and highlights critical considerations for future research in this domain.

END:VEVENT END:VCALENDAR