Universidade de São Paulo (USP)
Instituto de Ciências Matemáticas e de Computação (ICMC)
Faculdade de Medicina da Universidade de São Paulo (FMUSP)
Universidade Federal de São Carlos
Departamento de Gerontologia – Centro de Ciências Biológicas e da Saúde
Centro de Pesquisa e Desenvolvimento em Telecomunicações

New paradigms, scalability and applications of automated diagnosis and monitoring in aging: evaluating the relationship between healthy and cognitively impaired elderly

Principal Investigators
Computer Scientists
Sandra Maria Aluisio - GIAC/ICMC/USP (Natural Language Processing)
Moacir Antonelli Ponti ICMC/USP (Signal processing)
Renata Pontin de Mattos Fortes ICMC/USP (Web accessibility and Human-Computer Interaction) Fernando Oscar Runstein CPqD (Speech Processing Group)
Scientists in the Target Domain
Letícia Mansur - FMUSP (Neurolinguistics)
Paula Costa Castro UFSCar (Gerontology)

Associated Researchers
Andre Carvalho (ICMC), Diego Amancio (ICMC/USP), Osvaldo Novaes (IFSC/USP), Aline Cristina Martins Gratão (UFSCar), Marcos Hortes Nisihara Chagas (UFSCar), Lilian Cristine Hübner (PUC-RS), Denise Casatti (ICMC - Scientific Journalism), Lucas Ferreira de Lara (Coordenador do Programa Conte sua História do Museu da Pessoa), Sonia Brucki (FMUSP - Neuroscience), MARIA TERESA CARTHERY-GOULART (Universidade Federal do ABC-SP), ROCHELE PAZ FONSECA (PUCRS), MARIA ISABEL D´ÁVILA FREITAS (Universidade Federal de Santa Catarina - UFSC), LENISA BRANDÃO (UFRGS), THAIS HELENA MACHADO (PÓS-DOUTORANDA NO PROGRAMA SAÚDE DO ADULTO – UFMG)

International collaborations
Jed Meltzer (University of Toronto and Rotman Research Institute, Canada) and David Frohlich (University of Surrey, England)



Alzheimer’s disease (AD) is the most common form of dementia. It generates cognitive deficits serious enough to interfere in an individual’s daily life. It also grows in importance a less known syndrome, called Mild Cognitive Impairment (MCI); its most frequent type (the amnestic MCI) has the highest conversion rate to AD (15% per year, versus 1-2% of the total population). Recent studies have acknowledged the heterogeneity of AD and MCI, which increases the relevance of the analysis of other cognitive skills, such as language, besides evaluating memory. Language assessment sees in discursive production an attractive alternative, mainly in narratives, since it is a natural form of communication and it favors the observation of patient functionality in everyday life. However, analysis of speech samples when conducted by hand is a time-consuming, subjective work and difficult to become scalable to attend a large demand of elderly people. To address this problem, this project proposes a new computational infrastructure to support e-Science on clinical data to detect early signs of dementia, including Natural Language Processing tools and Machine Learning (ML) methods to process speech data of narratives. Moreover, the project will conduct a multicenter neuropsychological assessment on tablets and provide storage in databases in the cloud. In addition, the project will provide validated Portuguese-language versions of very well used batteries for cognitive assessments, such as UDS, retelling tests of ABCD and mini-ACE. As for methods, Deep Neural Networks will be pervasively used to analyze speech, text, signal, and image data. Complex Networks will provide a new representation for narratives in order to make available new metrics for machine learning methods to classify Controls, MCI and DA subjects. Finally, retelling tests are modeled as a variation of the Recognizing Textual Entailment task in which the recognition of propositions is automated by monolingual alignment methods, followed by a stage of text entailment decision.


Status: to be submitted to Fapesp Funding Agency. 

Students involved

Ongoing Research:

Leandro Borges dos Santos. ANAA-Dementia: Automated neuropsychological assessments for Brazilian citizens during their lifetime. (PhD) Google Scholarship/CNPq Scholarship.

Edresson Casanova. Reconhecimento automático de fala espontânea e transfer learning para cenários de fala comprometida. (PhD) CAPES Scholarship.

Finished Research:

Andre Luiz Verucci da CunhaCoh-Metrix-Dementia: análise automática de distúrbios de linguagem nas demências utilizando processamento de línguas naturais. (MsC) Fapesp Scholarship.

Marcos Treviso. Segmentação de sentenças e detecção de disfluências em narrativas transcritas de testes neuropsicológicos. (MsC) CNPq Scholarship.

Cíntia Matsuda Toledo. Análise de aspectos micro e macrolinguísticos da narrativa de indivíduos com doença de Alzheimer, comprometimento cognitivo leve e sem comprometimentos cognitivos. (Doutorado em Ciências da Reabilitação) - Faculdade Medicina da Universidade de São Paulo.


1. SANTOS, L. B. ; CORREA JR., E. A. ; OLIVEIRA JR., O. N. ; AMANCIO, D. R. ; MANSUR, L. L. ; ALUÍSIO, S. M. . Enriching Complex Networks with Word Embeddings for Detecting Mild Cognitive Impairment from Speech Transcripts. In: Annual Meeting of the Association for Computational Linguist 2017, 2017, Vancouver. Proceedings of the 55th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). Dublin, Ireland: Association for Computational Linguistics, 2017. v. 1. p. 1284-1296.

2. TOLEDO, CÍNTIA MATSUDA ; ALUÍSIO, Sandra Maria; DO SANTOS, LEANDRO BORGES ; BRUCKI, SONIA MARIA DOZZI ; TRÉS, EDUARDO STURZENEKER ; DE OLIVEIRA, MAIRA OKADA ; MANSUR, LETÍCIA LESSA. Analysis of macrolinguistic aspects of narratives from individuals with Alzheimer's disease, mild cognitive impairment, and no cognitive impairment. Alzheimer's & Dementia: Diagnosis, Assessment & Disease Monitoring, v. 10, p. 31-40, 2017.

3. TREVISO, M.; SHULBY, C.; ALUÍSIO, Sandra Maria. Sentence Segmentation in Narrative Transcripts from Neuropsychological Tests using Recurrent Convolutional Neural Networks. In: European Chapter of the Association for Computational Linguistics - EACL 2017, 2017, Valência. Proceedings of the 15th European Chapter of the Association for Computational Linguistics Conference, 2017. v. 1. p. 315-325.

4. TREVISO, M. ; SHULBY, C. ; ALUÍSIO, Sandra Maria. Evaluating Word Embeddings for Sentence Boundary Detection in Speech Transcripts. In: XI Brazilian Symposiumin Information and Human Language Technology - STIL 2017, 2017, Uberlândia. Proceedings of the XI Brazilian Symposiumin Information and Human Language Technology. Uberlândia, MG, Brazil, October 2?5, 2017.. Porto Alegre: Sociedade Brasileira de Computação, 2017. v. 1. p. 151-160.

5. ALUISIO, S. CUNHA, A. AND SCARTON, C. Evaluating progression of Alzheimer’s disease by regression and classification methods in a narrative language test in portuguese. In João Silva, Ricardo Ribeiro, Paulo Quaresma, André Adami, and António Branco, eds, 12th International Conference, PROPOR 2016, Tomar, Portugal, July 13-15, 2016, Proceedings, p. 109-114.

6. A Computational Tool for Automated Language Production Analysis Aimed at Dementia Diagnosis – Sandra Aluísio, Andre Cunha, Cintia Toledo and Carolina Scarton. 12th International Conference, PROPOR 2016, Tomar, Portugal, July 13-15, 2016, Demo Session Proceedings, 2016.

7. TOLEDO, C. M. ; CUNHA, A. ; ALUISIO, S. M. ; BAHIA, Valéria Santoro ; BRUCKI, S. M. D. ; MANSUR, LL . Coh-metrix dementia: automatic analysis of discourse in mild cognitive impairment and dementia. In: X Reunião de Pesquisadores em Doença de Alzheimer e Desordens Relacionadas, 2015, Fortaleza. Dementia & Neuropsychologia (Supplement 1). São Paulo: Área Visual Comunicação Gráfica Ltda, 2015. v. 9. p. 32-32.

8. Andre Luiz Verucci da Cunha, Lucilene Bender de Sousa, Leticia Lessa Mansur, Sandra Maria Aluísio: Automatic Proposition Extraction from Dependency Trees: Helping Early Prediction of Alzheimer's Disease from Narratives. CBMS 2015: p. 127-130, 2015.

9. TOLEDO, C.; CUNHA, A.; SCARTON, C.; ALUÍSIO, S. M. (2014) Automatic Classification of Written Descriptions by Healthy Adults: an Overview of the Application of Natural Language Processing and Machine Learning Techniques to Clinical Discourse Analysis. Dementia & Neuropsychologia, v. 8(3), p. 227-235, 2014.