Banca de QUALIFICAÇÃO: LAYS DIANY CRISANTO DA SILVA

Uma banca de QUALIFICAÇÃO de MESTRADO foi cadastrada pelo programa.
STUDENT : LAYS DIANY CRISANTO DA SILVA
DATE: 24/08/2026
TIME: 15:00
LOCAL: Online
TITLE:

Mobile Platform for Visualization and Automatic Segmentation of Cardiac Magnetic Resonance Images

KEY WORDS:

Mobile medical visualization, Gyroscope, Accelerometer, Volume ray marching, Cardiac imaging, Usability, NIfTI, Denoising Diffusion Probabilistic Models, nnU-Net, Synthetic data augmentation, ACDC.

PAGES: 75
BIG AREA: Ciências Exatas e da Terra
AREA: Ciência da Computação
SUBÁREA: Metodologia e Técnicas da Computação
SPECIALTY: Processamento Gráfico (Graphics)
SUMMARY:

Cardiovascular diseases are the leading cause of death worldwide (32\% of all global deaths in 2022~\parencite{WHO2025}, with associated healthcare costs projected to triple by 2050~\parencite{Kazi2024}). 3D visualization and automatic segmentation of cardiac exams can support diagnosis, but adoption is hindered by scarce annotated data and underexplored interaction interfaces on mobile devices.

This dissertation presents VisHeart3D, an ecosystem for mobile cardiac visualization on tablets, and investigates whether (H1) gyroscope and accelerometer interaction, combined with multi-touch gestures, reduces task time and/or improves usability (SUS) in 3D volume navigation for users without prior training, compared to touch-only mode, and whether (H2) synthetic pairs generated by a conditional DDPM increase the Dice score of an nnU-Net segmentor trained on ACDC data, relative to a real-data-only baseline.

Preliminary results confirm the platform's technical viability; the H1 pilot indicated that tilting the tablet complements touch during volumetric exploration; and the H2 real-only nnU-Net baseline has already been trained (validation Dice of 0.90), with the pipeline's remaining stages underway.

Contributions: the VisHeart3D ecosystem; an experimental protocol for evaluating inertial sensors in volumetric medical visualization; and a conditional DDPM data augmentation pipeline for cardiac segmentation in low-data regimes, a novel combination with nnU-Net and the ACDC dataset.

COMMITTEE MEMBERS:
Presidente - 2177445 - BRUNO MOTTA DE CARVALHO
Interna - 1350250 - ANNE MAGALY DE PAULA CANUTO
Externo ao Programa - 2497950 - SELAN RODRIGUES DOS SANTOS - UFRN