Skip to content

Latest commit

 

History

History
92 lines (68 loc) · 4.98 KB

README.md

File metadata and controls

92 lines (68 loc) · 4.98 KB

Schizophrenia prediction: use of Neuroimages and Artificial Intelligence Models.

Detection of schizophrenia based on tractographic techniques with fine tuning in machine learning.

Title repo tags date
Pitch Project Open In Google Neuroimages 2023-05-24
Dataset Open In Google UCLA 2023-05-24

project folders

[Tractografía]

  • data: Raw data for the project and images.
  • docs: Documentation, including Markdown.
  • results: Results, including checkpoints, hdf5 files, pickle files, as well as figures and tables.
  • scripts: Scripts (Python, bash, .ipynb notebooks).
  • src: Reusable Python modules for the project (imports).
  • tests: Tests for the code.

Project definition

Main Question

Is it possible to combine tractography analysis, fMRI and Machine Learning to create an artificial intelligence model in order to enhance schizophrenia prediction?

Background

Schizophrenia theories: Theory of Disconnection Syndrome

  • The theory of disconnection syndrome is a theory that explains the symptoms of schizophrenia as the result of disruptions in the normal integration of emotion, perception, and thought.

Previous works and literature: Use of only one method of analyses, Require additional validation afterwards

MRI: Now it’s only use for differential diagnosis and not for prediction

Tools

Methodology

  • Preprocessing, processing and tractography and fMRI using software DSI Studio and SPM
  • Creating and training AI model
  • Statistical analyses

Data

Objectives

In Spanish:

  • Diseñar un modelo de aprendizaje automático que detecte anomalías y mecanismos subyacentes a traves de la conectómica de la corteza prefrontal, luego clasifique en categorías, y finalmente prediga el diagnóstico de la enfermedad de esquizofrenia.

Específicamente se propone:

  1. Utilizar técnicas de tractografía en resonancias magnéticas estructurales y por tensor de difusión para capturar la conectividad estructural y transformar en dataset para aprendizaje automatico.
  • Evaluar las conexiones existentes del PFC:
    • Fascículo arqueado
    • Parietal frontal del cíngulo
    • SLF
    • Fascículo uncinado
    • Red de modo predeterminado
  1. Crear un script que separa los datos en entrenamiento y prueba, y que se pueda modificar el tamaño de la muestra.
  2. Comparar modelos generales de clasificación para determinar el mejor modelo.
  3. Comparar El conectoma de sujeto sano (grupo 1) con sujetos con diagnóstico de esquizofrenia (grupo 2)

An English:

  • Design a machine learning model that detects abnormalities and underlying mechanisms through connectomics of the prefrontal cortex, then classifies them into categories, and finally predicts the diagnosis of schizophrenia illness.

Specifically:

  1. Use tractography techniques in structural and diffusion tensor MRI to capture structural connectivity, also, and transform into dataset for machine learning.
  • Evaluate he existing connections of the PFC:
    • Arcuate Fasciculus
    • Cingulum Frontal Parietal
    • SLF
    • Uncinate Fasciculus
    • Default mode network
  1. Create a script that separates the data in training and test, and that can modify the size of the sample.
  2. Compare general classification models to determine the best model.
  3. Compare The connectome of healthy subject (group 1) with subjects with a diagnosis of schizophrenia (group 2)

Deliverables

  • A Github repository with codes and scripts to reproduce training and testing.
  • A jupyter notebook of the analysis codes and visualisations for comparing the results.
  • Documentation

Colaborators

H.Galván A.Boveda P.Koss S.Galván