Machine-learning-based biomarkers towards customization of robotic rehabilitation treatments for stroke patients
Autor(a) principal: | |
---|---|
Data de Publicação: | 2021 |
Tipo de documento: | Tese |
Idioma: | eng |
Título da fonte: | Biblioteca Digital de Teses e Dissertações da USP |
Texto Completo: | https://www.teses.usp.br/teses/disponiveis/55/55134/tde-24062021-160523/ |
Resumo: | Coupled with sensors, robotic devices for Stroke rehabilitation describe the motor behavior of patients as kinematic and kinetic data, which are underexplored in the data science and machine learning context, due to the time-consuming task of pursuing enough data volume. Moreover, the definition of biomarkers for a quantitative and deterministic assessment of patient progress remains an open problem in the literature. Four different studies were carried out aiming to address such an issue. We also propose a modular framework for organizing pieces of software for data analysis with the calculations of more than twenty robot-based metrics implemented. Our first study concerns a clinically-naive method for defining a region in the data space, alluding to a state of rehabilitation, based on the uncertainty in the classification of hemiparetic sides of chronic stroke patients. Our second study raised evidences that anodal tDCS may have a detrimental or maladaptive interaction with the affected hemisphere in patients with very severe upper-extremity impairments. Our third study correlates the implemented robot-based metrics with traditional clinical scales, so trained machine learning models can play the same role on a quantitative and deterministic way, eliminating the subjective nature from traditional evaluation methods. We found that having a model to estimate clinical scales only from one kind of robot (shoulder/elbow or wrist) is as good as combining data from both. We found in our fourth study evidences from a clinical perspective on the prediction of clinical outcomes of patients at early stages of the treatment. Our results indicate the possibility of improving the decision-making process by alerting, at the end of the second therapy session, when patients will potentially not present a significant response. The four projects here described enabled to push the state of the art towards the development of biomarkers to evaluate and track patients progress on rehabilitation robotics treatments; proposing standards for simplifying the data sharing; simplifying clinical studies with traditional statistical tools and our framework API and optimize the clinicians decision-making process towards impacting patients budget on rehabilitation treatments for optimizing quality of life. |
id |
USP_fcfeca668635ac30a6dc22539da61719 |
---|---|
oai_identifier_str |
oai:teses.usp.br:tde-24062021-160523 |
network_acronym_str |
USP |
network_name_str |
Biblioteca Digital de Teses e Dissertações da USP |
repository_id_str |
2721 |
spelling |
Machine-learning-based biomarkers towards customization of robotic rehabilitation treatments for stroke patientsBiomarcadores baseados em aprendizado de máquina para customização de tratamentos de reabilitação robótica para pacientes com AVCAcidente vascular cerebralAprendizado de máquinaBiomarcadoresBiomarkersMachine learningRehabilitation roboticsRobótica de reabilitaçãoStrokeCoupled with sensors, robotic devices for Stroke rehabilitation describe the motor behavior of patients as kinematic and kinetic data, which are underexplored in the data science and machine learning context, due to the time-consuming task of pursuing enough data volume. Moreover, the definition of biomarkers for a quantitative and deterministic assessment of patient progress remains an open problem in the literature. Four different studies were carried out aiming to address such an issue. We also propose a modular framework for organizing pieces of software for data analysis with the calculations of more than twenty robot-based metrics implemented. Our first study concerns a clinically-naive method for defining a region in the data space, alluding to a state of rehabilitation, based on the uncertainty in the classification of hemiparetic sides of chronic stroke patients. Our second study raised evidences that anodal tDCS may have a detrimental or maladaptive interaction with the affected hemisphere in patients with very severe upper-extremity impairments. Our third study correlates the implemented robot-based metrics with traditional clinical scales, so trained machine learning models can play the same role on a quantitative and deterministic way, eliminating the subjective nature from traditional evaluation methods. We found that having a model to estimate clinical scales only from one kind of robot (shoulder/elbow or wrist) is as good as combining data from both. We found in our fourth study evidences from a clinical perspective on the prediction of clinical outcomes of patients at early stages of the treatment. Our results indicate the possibility of improving the decision-making process by alerting, at the end of the second therapy session, when patients will potentially not present a significant response. The four projects here described enabled to push the state of the art towards the development of biomarkers to evaluate and track patients progress on rehabilitation robotics treatments; proposing standards for simplifying the data sharing; simplifying clinical studies with traditional statistical tools and our framework API and optimize the clinicians decision-making process towards impacting patients budget on rehabilitation treatments for optimizing quality of life.Acoplados a sensores, robôs para reabilitação de AVC descrevem o comportamento motor de pacientes como grandezas cinemáticas e dinâmicas, pouco exploradas no contexto de ciência de dados, devido à custosa tarefa de obter um volume significativo de dados. Além disso, a definição de biomarcadores para uma avaliação mais confiável da evolução do paciente permanece um problema aberto na literatura. Quatro estudos diferentes foram conduzidos com o objetivo de abordar tal questão. É proposta também uma ferramenta modular para organizar programas para análise de dados, com cálculos de mais de vinte métricas implementadas. O primeiro estudo consiste em um método puramente baseado em dados para definir uma região no espaço de dados, aludindo a um estado de reabilitação, baseado na incerteza da classificação de lados hemiparéticos de pacientes com AVC crônico. Nosso segundo estudo levantou evidências de que a tDCS anódica pode ter uma interação desvantajosa com o hemisfério afetado em pacientes com déficis muito severos dos membros superiores. O terceiro estudo correlaciona as métricas implementadas com escalas clínicas tradicionais, de forma que os modelos de aprendizado de máquina treinados possam desempenhar o mesmo papel de forma quantitativa e determinística, eliminando a natureza subjetiva dos métodos tradicionais de avaliação. Descobrimos que utilizar o modelo treinado com dados de apenas de um tipo de robô (ombro/cotovelo ou pulso) para estimar escalas clínicas é tão eficiente quanto combinar dados de ambos. Encontramos no quarto estudo evidências, sob uma perspectiva clínica, de um potencial de previsão de resultados clínicos de pacientes em estágios iniciais do tratamento. Nossos resultados indicam a possibilidade de melhorar o processo de tomada de decisão alertando, ao final da segunda sessão, se o paciente potencialmente não apresentará uma resposta significativa à terapia. Os projetos aqui descritos avançam o estado da arte no desenvolvimento de biomarcadores para avaliar e acompanhar o progresso do paciente em terapia robótica; propõem padrões a fim de simplificar o compartilhamento de dados; simplificam os estudos clínicos com amparo estatístico, bem como a ferramenta proposta, e otimizam a tomada de decisão clínica, impactando o orçamento em tratamentos de reabilitação e otimizando recursos do paciente para uma melhor qualidade de vida.Biblioteca Digitais de Teses e Dissertações da USPDelbem, Alexandre Cláudio BotazzoMoretti, Caio Benatti2021-03-31info:eu-repo/semantics/publishedVersioninfo:eu-repo/semantics/doctoralThesisapplication/pdfhttps://www.teses.usp.br/teses/disponiveis/55/55134/tde-24062021-160523/reponame:Biblioteca Digital de Teses e Dissertações da USPinstname:Universidade de São Paulo (USP)instacron:USPLiberar o conteúdo para acesso público.info:eu-repo/semantics/openAccesseng2021-06-24T22:11:02Zoai:teses.usp.br:tde-24062021-160523Biblioteca Digital de Teses e Dissertaçõeshttp://www.teses.usp.br/PUBhttp://www.teses.usp.br/cgi-bin/mtd2br.plvirginia@if.usp.br|| atendimento@aguia.usp.br||virginia@if.usp.bropendoar:27212021-06-24T22:11:02Biblioteca Digital de Teses e Dissertações da USP - Universidade de São Paulo (USP)false |
dc.title.none.fl_str_mv |
Machine-learning-based biomarkers towards customization of robotic rehabilitation treatments for stroke patients Biomarcadores baseados em aprendizado de máquina para customização de tratamentos de reabilitação robótica para pacientes com AVC |
title |
Machine-learning-based biomarkers towards customization of robotic rehabilitation treatments for stroke patients |
spellingShingle |
Machine-learning-based biomarkers towards customization of robotic rehabilitation treatments for stroke patients Moretti, Caio Benatti Acidente vascular cerebral Aprendizado de máquina Biomarcadores Biomarkers Machine learning Rehabilitation robotics Robótica de reabilitação Stroke |
title_short |
Machine-learning-based biomarkers towards customization of robotic rehabilitation treatments for stroke patients |
title_full |
Machine-learning-based biomarkers towards customization of robotic rehabilitation treatments for stroke patients |
title_fullStr |
Machine-learning-based biomarkers towards customization of robotic rehabilitation treatments for stroke patients |
title_full_unstemmed |
Machine-learning-based biomarkers towards customization of robotic rehabilitation treatments for stroke patients |
title_sort |
Machine-learning-based biomarkers towards customization of robotic rehabilitation treatments for stroke patients |
author |
Moretti, Caio Benatti |
author_facet |
Moretti, Caio Benatti |
author_role |
author |
dc.contributor.none.fl_str_mv |
Delbem, Alexandre Cláudio Botazzo |
dc.contributor.author.fl_str_mv |
Moretti, Caio Benatti |
dc.subject.por.fl_str_mv |
Acidente vascular cerebral Aprendizado de máquina Biomarcadores Biomarkers Machine learning Rehabilitation robotics Robótica de reabilitação Stroke |
topic |
Acidente vascular cerebral Aprendizado de máquina Biomarcadores Biomarkers Machine learning Rehabilitation robotics Robótica de reabilitação Stroke |
description |
Coupled with sensors, robotic devices for Stroke rehabilitation describe the motor behavior of patients as kinematic and kinetic data, which are underexplored in the data science and machine learning context, due to the time-consuming task of pursuing enough data volume. Moreover, the definition of biomarkers for a quantitative and deterministic assessment of patient progress remains an open problem in the literature. Four different studies were carried out aiming to address such an issue. We also propose a modular framework for organizing pieces of software for data analysis with the calculations of more than twenty robot-based metrics implemented. Our first study concerns a clinically-naive method for defining a region in the data space, alluding to a state of rehabilitation, based on the uncertainty in the classification of hemiparetic sides of chronic stroke patients. Our second study raised evidences that anodal tDCS may have a detrimental or maladaptive interaction with the affected hemisphere in patients with very severe upper-extremity impairments. Our third study correlates the implemented robot-based metrics with traditional clinical scales, so trained machine learning models can play the same role on a quantitative and deterministic way, eliminating the subjective nature from traditional evaluation methods. We found that having a model to estimate clinical scales only from one kind of robot (shoulder/elbow or wrist) is as good as combining data from both. We found in our fourth study evidences from a clinical perspective on the prediction of clinical outcomes of patients at early stages of the treatment. Our results indicate the possibility of improving the decision-making process by alerting, at the end of the second therapy session, when patients will potentially not present a significant response. The four projects here described enabled to push the state of the art towards the development of biomarkers to evaluate and track patients progress on rehabilitation robotics treatments; proposing standards for simplifying the data sharing; simplifying clinical studies with traditional statistical tools and our framework API and optimize the clinicians decision-making process towards impacting patients budget on rehabilitation treatments for optimizing quality of life. |
publishDate |
2021 |
dc.date.none.fl_str_mv |
2021-03-31 |
dc.type.status.fl_str_mv |
info:eu-repo/semantics/publishedVersion |
dc.type.driver.fl_str_mv |
info:eu-repo/semantics/doctoralThesis |
format |
doctoralThesis |
status_str |
publishedVersion |
dc.identifier.uri.fl_str_mv |
https://www.teses.usp.br/teses/disponiveis/55/55134/tde-24062021-160523/ |
url |
https://www.teses.usp.br/teses/disponiveis/55/55134/tde-24062021-160523/ |
dc.language.iso.fl_str_mv |
eng |
language |
eng |
dc.relation.none.fl_str_mv |
|
dc.rights.driver.fl_str_mv |
Liberar o conteúdo para acesso público. info:eu-repo/semantics/openAccess |
rights_invalid_str_mv |
Liberar o conteúdo para acesso público. |
eu_rights_str_mv |
openAccess |
dc.format.none.fl_str_mv |
application/pdf |
dc.coverage.none.fl_str_mv |
|
dc.publisher.none.fl_str_mv |
Biblioteca Digitais de Teses e Dissertações da USP |
publisher.none.fl_str_mv |
Biblioteca Digitais de Teses e Dissertações da USP |
dc.source.none.fl_str_mv |
reponame:Biblioteca Digital de Teses e Dissertações da USP instname:Universidade de São Paulo (USP) instacron:USP |
instname_str |
Universidade de São Paulo (USP) |
instacron_str |
USP |
institution |
USP |
reponame_str |
Biblioteca Digital de Teses e Dissertações da USP |
collection |
Biblioteca Digital de Teses e Dissertações da USP |
repository.name.fl_str_mv |
Biblioteca Digital de Teses e Dissertações da USP - Universidade de São Paulo (USP) |
repository.mail.fl_str_mv |
virginia@if.usp.br|| atendimento@aguia.usp.br||virginia@if.usp.br |
_version_ |
1809091043650437120 |