Quantifying WMH and brain volumes in heterogeneous clinical and low-field portable MRI
Built the first method that can simultaneously segment brain ROIs and WMH in scans of any resolution and contrast, including pMRI.
View GitHub repo
×
Deprecated: This page is outdated since 2023.
×
Info: New portfolio web design is available at https://lasopablo.github.io/vcard_portfolio/
Projects
2023
Adversarial Attacks and Defenses on Neural Networks for Image Classification
The project involves creating a Neural Network for classification, executing an adversarial attack, implementing a defense, and analyzing and comparing the attack and defense from a theoretical and numerical perspective. We end up building an RNN-like architecure.
View on Google Collab
Music Genre Prediction on Big Data
Explored the application of machine learning techniques for predicting music genre using the Million Songs Dataset. After a statistical analysis, Decision Tree, Random Forest, Linear Regression, and Random Forest algorithms are applied.
View on Google Collab
Lyrics Based Song Genre Classification
The project aims to classify songs into different genres based on their lyrics, exploring the accuracy and reliability of NLP tools for genre prediction, using machine learning classifiers and three different ways of embedding text.
View on Google Collab
Comparison of Traditional Bag of Words Approach to Deep Learning for Image Retrieval
The study compared various Information Retrieval systems for Image Retrieval, including traditional Bag of Words approach and state-of-the-art Convolutional Neural Networks.
View on Google Collab
2022
Image Processing Server at KI
Implemented an Image Processing Management server at Karolinska Institutet (known for awarding the Nobel Prize in Medicine) to automate all Image Processing steps as a streamlined pipeline.
Comparing Machine Learning Algorithms For Link Prediction
Worked on a method to predict links between pairs of nodes in networks using machine learning algorithms and cosine similarity on node embeddings calculated by DeepWalk. Grid search is used to optimize the parameters of the models and compare their results.
Studying the Most Relevant Risk Factors for Heart Disease
We study the possibility of developing a machine learning model that is able to early predict an heart failure case. Moreover, the study shows which are the most relevant risks for heart disease.
View on Google Collab
2021
Computer-aided Decision (CAD) System
Implemented a CAD system for General Electric at Quirón hospital (Madrid, Spain) as part of my Bachelor's thesis. This system uses a ML- and Deep Learning-based algorithm to detect possible cancerous areas in prostate MR images, classify them as either a "clinically significant" or "clinically insignificant" finding. When identified as significant, it will also suggest a PI-RADS score. Finally, the CAD system generates a report with key statistics from the finding and image, to assist the physician in finding abnormal parameters.
MR Imaging and Instrumentation
Contributed to a tabletop MRI replication project, led by LAIMBIO (Madrid, Spain) in collaboration with the Martinos Center (MGH, Harvard Medical School) in Boston, MA, USA. Edited and expanded circuit and software designs to aid in the creation of a replica tabletop MRI scanner.
Deep Learning-based classifier in LUS
Developed two Deep Learning-based Image Recognition systems, incorporating Convolutional Neural Network (CNN) architecture and leveraging Transfer Learning. The goal was to recognize lung pathologies, namely, pneumonia and COVID-19, in Lung Ultrasound (US) images and videos.
View on Github repo
Statistical Analysis on Breast Cancer
In a small project for a course on AI and Inference, a Statistical Study was conducted on a datatset for Breast Cancer.
View on Google Collab
2020
Signal Processing and Anthropology study
The project is divided into (1) an EMG and (2) Physiological Signal Processing report; and (3) an Anthropology study, with a strong MATLAB-based coding component.
View on Github repo
2019
Electromyography
Built an OpAmp-based circuit for EMG activity detection and noise cancellation. The circuit was designed for EMG signal processing in an EMG study. EMG Signal Processing and Analysis was later performed.
Mute people able to speak through a myo-armband!
Developed on a myo-armband API to use muscle signals and arm/hand movements to articulate words.
2018
Server
Designed a simple server that accepts requests from a client and performs the appropiate query to retrieve the informataion from the FDA/Genius API.