Amelia Jiménez-Sánchez on DASYA https://dasya.itu.dk/supervisor/amelia-jimenez-sanchez/ Recent content in Amelia Jiménez-Sánchez on DASYA Hugo -- gohugo.io en-us © DASYA, IT University of Copenhagen Tue, 07 May 2024 00:00:00 +0000 Debiasing medical image datasets to improve machine learning robustness and fairness https://dasya.itu.dk/for-students/proposals/archive/ajs-debias/ Tue, 07 May 2024 00:00:00 +0000 https://dasya.itu.dk/for-students/proposals/archive/ajs-debias/ It has been observed that deep learning models are able to identify patient characteristics such as age, sex, and self-reported race with high accuracy from medical images such as chest x-ray recordings, even when medical doctors cannot. This raises the potential for such models to learn to (falsely) diagnose patients of different demographics differently, even if they present with the same disease characteristics. This may happen because, for example, some groups may tend to be misdiagnosed or underdiagnosed in the datasets using which such models are trained. Concept Bottleneck Models to detect hidden features or avoid memorization https://dasya.itu.dk/for-students/proposals/archive/ajs-concept-bottleneck-models/ Thu, 11 May 2023 00:00:00 +0000 https://dasya.itu.dk/for-students/proposals/archive/ajs-concept-bottleneck-models/ Concept Bottleneck Models [1] are designed to leverage high-level concepts. They revisit the classic idea of first predicting concepts that are providing at training time, and then using these concepts to predict the label. By construction, it is possible to intervene on these concept bottleneck models by editing their predicted concept values and propagating these changes to the final prediction. The goal of this project is to leverage the concept block to detect possible spurious correlations (shortcuts) [2,3], memorization [4] or unfair predictions [5]. Medical Visual Question Answering https://dasya.itu.dk/for-students/proposals/archive/ajs-medical-vqa/ Thu, 11 May 2023 00:00:00 +0000 https://dasya.itu.dk/for-students/proposals/archive/ajs-medical-vqa/ A medical Visual Question Answering (VQA) system can provide meaningful references for both doctors and patients during the treatment process. Different from normal images, a learning setting with medical images is more challenging due limited amounts of data, class-imbalance and the presence of label noise for diagnosis tasks. Moreover, little attention is paid to how the images and meta-data is obtained for medical datasets. Recent studies show that even with high performance on the existing data, algorithms can learn “shortcuts” like visibility of medical tools and fail to generalize.