Artificial Inteligence in Biomedicine Group (ArBio) Projects
Our research areas are focused in: image processing, computer vision and machine learning and deep learning.
In particular, our published works cover medical imaging (Rx, CT, MRI, PET), neuroimaging (MRI, DTI, fMRI) and photomicrography.
Also, we have addressed topics such as segmentation and detection, video tracking,
underwater monitoring, 3D reconstruction and anomaly detection.
Financial projects
Aprendizaje computacional multimodal para el análisis de imágenes y datos médicos.
Leader: Dr. Jorge Luis Pérez (IIMAS-Yucatán).
Sponsor: PAPIIT UNAM IA100924.
Period: January 2024 – December 2026.
Team:
Dra. Nidiyare Hevia Montiel.
Aplicación de técnicas de procesamiento de imágenes, visión computacional y aprendizaje automático
en el estudio y diagnóstico por imagenología médica de la infección experimental con Tripanosoma cruzi.
Leader: Dra. Nidiyare Hevia Montiel.
Sponsor: PAPIIT UNAM IT101422.
Period: January 2022 – December 2024.
Team:
Dr. Jorge Luis Pérez (IIMAS-Yucatán),
Dr. Antonio Neme (IIMAS – Yucatán),
Dra. Paulina Haro Álvarez (CIR – UADY),
Dra. Blanca Hilda Vázquez Gómez,
Mtro. José Leonardo Guillermo Cordero (Facultad de Medicina Veterinaria y Zootecnia, UADY)
Aplicación de técnicas de aprendizaje computacional en imagenología para el estudio y
análisis histopatológico de la infección experimental con Trypanosoma cruzi.
Leader: Dra. Nidiyare Hevia Montiel.
Sponsor: PAPIIT UNAM IT100220.
Period: January 2020 – December 2021.
Team:
Dra. Paulina Haro Álvarez (IIM – UAB),
Dr. Jorge Luis Pérez (IIMAS-Yucatán),
Dr. Antonio Neme (IIMAS – Yucatán).
Diseño e implementación de algoritmos de aprendizaje computacional para el análisis de imágenes y datos médicos multimodales.
Leader: Dr. Jorge Luis Pérez (IIMAS-Yucatán).
Sponsor: PAPIIT UNAM IA104622.
Period: January 2022 – December 2023.
Team:
Dra. Nidiyare Hevia Montiel (IIMAS – Yucatán).
Diseño e implementación de nuevos métodos de registro de imágenes médicas usando técnicas de aprendizaje maquinal.
Leader: Dr. Jorge Luis Pérez (IIMAS-Yucatán)
Sponsor: PAPIIT UNAM IA102920.
Period: January 2020 – December 2021.
Team:
Dra. Nidiyare Hevia Montiel (IIMAS – Yucatán). More information
The main contribution of this project was the design of point cloud registration algorithms using
computational learning algorithms such as neural networks or random forest. These algorithms have
various applications in medical image processing. The algorithms developed focused on: 1. Study of diseases such as Parkinson's and Alzheimer's. Multimodal analysis schemes of
PET/CT/MRI/DTI brain images were designed to segment brain substructures,
register the various neuroimaging modalities, extract metabolic indicators and design
classification algorithms between control subjects and those with some pathology.
These projects arose from the needs of the National Institute of Neurology and Neurosurgery.
The idea was to find multimodal indicators of these diseases.
As main results, it was found that during Parkinson's disease the
indicators related to metabolic changes are predominant compared to
morphological changes. On the other hand, for Alzheimer's and Cognitive Impairment,
it was found that the main changes were morphological and diffusion, but not metabolic.
This is due to the degradation of brain tissue during the disease. 2. Registration of depth and tomography images for the design of transtibial prostheses.
Algorithms are being designed to align the surface of the residual limb with computed
tomography studies. This has the advantage of updating the external shape information
while preserving internal structures such as bones, tendons, etc. The purpose has been
to optimize costs and avoid continuous radiation of participants in tomography studies.
This helps to have 3D reconstructions of the residual limb useful at various stages
of prosthesis design. 3. A data augmentation algorithm has been designed using rigid and deformable
registration algorithms, a left ventricular segmentation algorithm and a left ventricular
wall flow or deformation estimation algorithm, all using deep neural networks.
These algorithms have been implemented in echocardiography images of control subjects
and those with altered ejection fraction. The contribution lies in the study of the
changes or deformations of the ventricular walls during a disease that affects
the morphology of the heart. 4. Registration of multiple fetal brains on ultrasound images. This is important
for removing occlusions or shadow artifacts during acquisitions.
Fetal biometry studies consider brain indicators such as biparietal diameter
or cranial circumference, which are indicators of fetal health status.
The contribution is the generation of new fetal brain studies from different
ultrasound projections previously recorded and fused. 5. Training of human resources. At the beginning of the project there was
only one integrated student, and during these two years a total of 8
students have participated in the project directly. All of them have
completed their studies or are in the final stages or waiting to be
assigned a date for their degree exam.
These works have been supported by the PAPIIT-UNAM programs IA102920, IA104622, IT100220, IT101422 and IA100924.
We are grateful for the funding received.
