Banca de DEFESA: WALTER SOARES BARBOSA ROCHA FILHO

Uma banca de DEFESA de DOUTORADO foi cadastrada pelo programa.
STUDENT : WALTER SOARES BARBOSA ROCHA FILHO
DATE: 10/09/2026
TIME: 09:00
LOCAL: Laboratório de Inovação Tecnológica em Saúde (LAIS)
TITLE:

AMALIA: ARTIFICIAL INTELLIGENCE-BASED PLATFORM FOR AUTOMATED MEASUREMENT, ANALYSIS AND REPORTING OF ELECTROCARDIOGRAMS IN SUPPORT OF MEDICAL DIAGNOSIS

KEY WORDS:

Artificial intelligence; Computer-aided diagnosis; Electrocardiogram; Machine learning; National Telediagnosis Platform (PNTD); Natural language processing; Public health; Telediagnosis.

PAGES: 100
BIG AREA: Engenharias
AREA: Engenharia Biomédica
SUMMARY:

Cardiovascular diseases are the leading cause of death worldwide and present a particular challenge in Brazil, where inequality in access to cardiological diagnosis is compounded by a shortage of specialists in peripheral and rural regions and by the overload of the public health system. In this context, the electrocardiogram is the reference examination for the screening and monitoring of cardiac conditions, but its interpretation is a cognitively complex task subject to interobserver. This thesis proposes AMALIA: an Artificial Intelligence-Based Assistant for Automated Measurement, Analysis and Reporting of Electrocardiograms, a modular platform for electrocardiogram analysis support developed on data from the Brazilian National Telediagnosis Platform of the Ministry of Health. The platform comprises three integrated modules: CardIA, responsible for processing the ECG signal in JSON format, automatically extracting morphological parameters by lead, and generating a preliminary report with confidence grading; AnalytIA, which operates on the structured examination database applying machine learning techniques for epidemiological analysis, clinical profile clustering, association rule mining, and individualized risk prediction; and AuditIA, which closes the cycle by comparing automatically generated reports with physician reference reports through natural language processing, computing concordance metrics, and guiding continuous algorithm calibration. The platform was designed to operate with low-cost artificial intelligence models, preferably running locally, eliminating dependency on foreign proprietary services and supporting compliance with Brazil's General Data Protection Law. The results obtained demonstrate the technical feasibility of the solution and its capacity to support cardiologists in producing faster and more standardized reports, with the potential to reduce diagnostic queues and broaden equity in access to cardiological care within Brazil's Unified Health System.

COMMITTEE MEMBERS:
Externo à Instituição - ANTONIO HIGOR FREIRE DE MORAIS - IFRN
Externo à Instituição - FELIPE RICARDO DOS SANTOS FERNANDES - ESEnfC
Externo à Instituição - HÊIDER AURÉLIO PINTO - UFBA
Externa à Instituição - JANAÍNA LUANA RODRIGUES DA SILVA VALENTIM - IFRN
Externo à Instituição - JORGE HENRIQUES - UC
Externa ao Programa - 2562782 - KARILANY DANTAS COUTINHO - UFRNInterno - 1153006 - LUIZ AFFONSO HENDERSON GUEDES DE OLIVEIRA
Presidente - 2488270 - RICARDO ALEXSANDRO DE MEDEIROS VALENTIM