As the application of deep learning in clinical practice is becoming more and more popular, such approaches are usually based on black-box models, making it difficult to explain why a classifier gave a specific output. End-to-end deep learning models might allow to interpret the results with dedicated visualization techniques.
In this regard, thesis projects on such topic can help to create a new deep learning approach for early medical diagnosis in different clinical scenarios, optimizing existing methodologies and creating end-to-end models.
Specifically, there are thesis projects available on four main scenarios:
- atrial fibrillation (with ECG signals)
- abnormal heart sounds (with Phonocardiagrams)
- pulmonary cancer histotypes (from CT scans)
- normal pressure hydrocephalus (from RMI)
The candidate is expected to review the state of the art of deep learning approaches in these scenarios and develop classification models by implementing evaluation and comparisons.
The thesis will be developed at the Information Engineering Department of the “Università Politecnica delle Marche”.
Involed tools and technologies: Keras, TensorFlow, Google Colab, Python, and similar.