ALS is a neurodegenerative disease that causes progressive loss of motor skills, and leads to difficulties in breathing, speaking, swallowing and eventually death, usually in a few years. Despite the lack of treatments, interventions such as non-invasive mechanical ventilation and percutaneous endoscopic gastrostomy can be made to prolong life expectancy when needed. Hence it would be clinically relevant to predict the patients’ need of such interventions. To this aim, the Intelligent Disease Progression Prediction challenge was organized, in which partecipants were tasked with developing new methods for risk and time-to-event prediction based on demographical and clinical features. Specifically, the challenge tasks consisted of predicting multiple competing risks, all related to ALS disease progression. We employ several machine learning methods generally applied to survival analysis and classification tasks, some of which are specialized for handling competing risks. All models were optimized through a cross-validation procedure and finally evaluated on an internal test set. The three best performing methods, namely Deep Survival Machines, Gradient boosted regression trees and Time-Aware Classifier Ensemble were selected and submitted to the IDPP challenge at CLEF 2022. The results of the competition showed that our methods achieve on average a c-index of ∼ 0.70 and ∼ 0.74, using data at time zero and up to six months, respectively. © 2022 Copyright for this paper by its authors.

Multi-Event Survival Prediction for Amyotrophic Lateral Sclerosis

Corrado Pancotti
;
Giovanni Birolo;Tiziana Sanavia;Cesare Rollo;Piero Fariselli
2022-01-01

Abstract

ALS is a neurodegenerative disease that causes progressive loss of motor skills, and leads to difficulties in breathing, speaking, swallowing and eventually death, usually in a few years. Despite the lack of treatments, interventions such as non-invasive mechanical ventilation and percutaneous endoscopic gastrostomy can be made to prolong life expectancy when needed. Hence it would be clinically relevant to predict the patients’ need of such interventions. To this aim, the Intelligent Disease Progression Prediction challenge was organized, in which partecipants were tasked with developing new methods for risk and time-to-event prediction based on demographical and clinical features. Specifically, the challenge tasks consisted of predicting multiple competing risks, all related to ALS disease progression. We employ several machine learning methods generally applied to survival analysis and classification tasks, some of which are specialized for handling competing risks. All models were optimized through a cross-validation procedure and finally evaluated on an internal test set. The three best performing methods, namely Deep Survival Machines, Gradient boosted regression trees and Time-Aware Classifier Ensemble were selected and submitted to the IDPP challenge at CLEF 2022. The results of the competition showed that our methods achieve on average a c-index of ∼ 0.70 and ∼ 0.74, using data at time zero and up to six months, respectively. © 2022 Copyright for this paper by its authors.
2022
Conference and Labs of the Evaluation Forum
Bologna
05/09/2022
CEUR Workshop Proceedings - 2022 Conference and Labs of the Evaluation Forum
Faggioli G, Ferro N, Hanbury A, Potthast M
3180
1269
1276
Amyotrophic Lateral Sclerosis; Machine Learning; Survival Prediction
Corrado Pancotti, Giovanni Birolo, Tiziana Sanavia, Cesare Rollo, Piero Fariselli
File in questo prodotto:
File Dimensione Formato  
paper-93.pdf

Accesso aperto

Tipo di file: PDF EDITORIALE
Dimensione 953.03 kB
Formato Adobe PDF
953.03 kB Adobe PDF Visualizza/Apri

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/2318/1877762
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus 2
  • ???jsp.display-item.citation.isi??? ND
social impact