Pedro R. A. S. Bassi,
M.Sc.
 

Johns Hopkins University - Baltimore, MD, USA

• 08/2024-02/2025 (expected): Visiting Ph.D. Student Ph.D. Student | Deep learning for 3D medical image analysis

  • Organized and led Touchstone (NeurIPS, 2024), a large-scale benchmark for organ segmentation in CT volumes. It involved the collaboration of 14 international research teams and focused on out-of-distribution generalization

  • Studied demographic biases and generalization in 3D medical segmentation

  • Participated in the creation of AbdomenAtlas, the largest fully-annotated abdominal CT dataset

  • Used large vision-language models to improve per-voxel organ annotations

  • Used large vision-language models to generate radiology reports for CT scans

​

University of Bologna - Bologna, BO, Italy

• 11/2021-11/2025 (expected): Ph.D. Researcher | Deep Learning Trustworthiness, Explainability and Distributional Robustness

  • Introduced the optimization of Layer-wise Relevance Propagation to control deep classifiers and improve out-of-distribution generalization​

  • Introduced and investigated methods to improve the inter-institution generalizability of deep X-ray classifiers

  • Studied shortcut learning in computer vision

  • Elaborated novel solutions to the problem of spurious correlations in the backgrounds of natural and biomedical images

​​​

Campinas State University (UNICAMP) - Campinas, SP, Brazil

• 08/2019-08/2021: M.Sc. Researcher | Deep Learning for Image Recognition in Brain-Computer Interfaces and X-ray Classification​

  • Studied the utilization of deep neural networks for AI-assisted diagnosis in chest X-rays​

  • Used explanation techniques to interpret the decisions of deep X-ray classifiers and compare them to radiologists' reasoning

  • Identified spurious correlations in early Covid-19 chest X-ray datasets

  • Evaluated the importance of lung segmentation for the inter-institutional generalizability of deep X-ray classifiers

  • Introduced the combination of filter banks and deep neural networks to improve the speed, portability and accuracy of brain-computer interfaces

  • Introduced the utilization of the SpecAugment data augmentation technique for the classification of EEG spectrograms in brain-computer interfaces​​

​

Campinas State University (UNICAMP) - Campinas, SP, Brazil

• 08/2018-07/2019: Undergraduate Researcher | Applications of deep learning in SSVEP (steady state visually evoked potential) based brain-computer interfaces

  • Utilized deep learning to classify spectrograms of EEG signals in SSVEP-based brain-computer interfaces

  • Introduced the use of deep triplet neural networks in brain-computer interfaces

​​​

Campinas State University (UNICAMP) - Campinas, SP, Brazil

• 08/2017-07/2018: Undergraduate Researcher | ​Fundamentals of deep neural networks

  • Studied machine learning and deep learning, encompassing fundamental concepts in linear algebra, information theory and probability​

  • Studied the book Deep Learning, written by Ian Goodfellow, Yoshua Bengio, and Aaron Courville

  • Implemented deep neural networks in Python, employing PyTorch and Tensorflow

  • Studied regularization and overfitting in deep neural networks