Tânia Pereira

Abstract
Nuclear-derived morphological features and biomarkers provide relevant insights regarding the tumour microenvironment, while also allowing diagnosis and prognosis in specific cancer types. However, manually annotating nuclei from the gigapixel Haematoxylin and Eosin (H&E)-stained Whole Slide Images (WSIs) is a laborious and costly task, meaning automated algorithms for cell nuclei instance segmentation and classification could alleviate the workload of pathologists and clinical researchers and at the same time facilitate the automatic extraction of clinically interpretable features for artificial intelligence (AI) tools. But due to high intra- and inter-class variability of nuclei morphological and chromatic features, as well as H&Estains susceptibility to artefacts, state-of-the-art algorithms cannot correctly detect and classify instances with the necessary performance. In this work, we hypothesize context and attention inductive biases in artificial neural networks (ANNs) could increase the performance and generalization of algorithms for cell nuclei instance segmentation and classification. To understand the advantages, use-cases, and limitations of context and attention-based mechanisms in instance segmentation and classification, we start by reviewing works in computer vision and medical imaging. We then conduct a thorough survey on context and attention methods for cell nuclei instance segmentation and classification from H&E-stained microscopy imaging, while providing a comprehensive discussion of the challenges being tackled with context and attention. Besides, we illustrate some limitations of current approaches and present ideas for future research. As a case study, we extend both a general (Mask-RCNN) and a customized (HoVer-Net) instance segmentation and classification methods with context- and attention-based mechanisms and perform a comparative analysis on a multicentre dataset for colon nuclei identification and counting. Although pathologists rely on context at multiple levels while paying attention to specific Regions of Interest (RoIs) when analysing and annotating WSIs, our findings suggest translating that domain knowledge into algorithm design is no trivial task, but to fully exploit these mechanisms in ANNs, the scientific understanding of these methods should first be addressed.

Abstract
The large gap between the generalization level of state-of-the-art machine learning and human learning systems calls for the development of artificial intelligence (AI) models that are truly inspired by human cognition. In tasks related to image analysis, searching for pixel-level regularities has reached a power of information extraction still far from what humans capture with image-based observations. This leads to poor generalization when even small shifts occur at the level of the observations. We explore a perspective on this problem that is directed to learning the generative process with causality-related foundations, using models capable of combining symbolic manipulation, probabilistic reasoning, and pattern recognition abilities. We briefly review and explore connections of research from machine learning, cognitive science, and related fields of human behavior to support our perspective for the direction to more robust and human-like artificial learning systems.

Abstract
Significant medical image visualization and annotation tools, tailored for clinical users, play a crucial role in disease diagnosis and treatment. Developing algorithms for annotation assistance, particularly machine learning (ML)-based ones, can be intricate, emphasizing the need for a user-friendly graphical interface for developers. Many software tools are available to meet these requirements, but there is still room for improvement, making the research for new tools highly compelling. The envisioned tool focuses on navigating sequences of DICOM images from diverse modalities, including Magnetic Resonance Imaging (MRI), Computed Tomography (CT) scans, Ultrasound (US), and X-rays. Specific requirements involve implementing manual annotation features such as freehand drawing, copying, pasting, and modifying annotations. A scripting plugin interface is essential for running Artificial Intelligence (AI)-based models and adjusting results. Additionally, adaptable surveys complement graphical annotations with textual notes, enhancing information provision. The user evaluation results pinpointed areas for improvement, including incorporating some useful functionalities, as well as enhancements to the user interface for a more intuitive and convenient experience. Despite these suggestions, participants praised the application's simplicity and consistency, highlighting its suitability for the proposed tasks. The ability to revisit annotations ensures flexibility and ease of use in this context.

Abstract
The early and accurate detection and the grading characterization of brain cancer will generate a positive impact on the treatment plan of those patients. AI-based models can help analyze the Magnetic Resonance Imaging (MRI) to make an initial assessment of the tumor grading. The objective of this work was to develop an Al-based model to classify the grading of the tumor using the MRI. Two regions of interest were explored, with several levels of complexity for the neural network architecture, and Iwo strategies to deal with Unbalanced data. The best results were obtained for the most complex architecture (Resnet50) with a combination of weighted random sampler and data augmentation achieving a balanced accuracy of 62.26%. This work confirmed that complex problems required a more dense neural network and strategies to deal with the unbalanced data.

Abstract
Recent studies have shown that the microbiome can impact cancer development, progression, and response to therapies suggesting microbiome-based approaches for cancer characterization. As cancer-related signatures are complex and implicate many taxa, their discovery often requires Machine Learning approaches. This review discusses Machine Learning methods for cancer characterization from microbiome data. It focuses on the implications of choices undertaken during sample collection, feature selection and pre-processing. It also discusses ML model selection, guiding how to choose an ML model, and model validation. Finally, it enumerates current limitations and how these may be surpassed. Proposed methods, often based on Random Forests, show promising results, however insufficient for widespread clinical usage. Studies often report conflicting results mainly due to ML models with poor generalizability. We expect that evaluating models with expanded, hold-out datasets, removing technical artifacts, exploring representations of the microbiome other than taxonomical profiles, leveraging advances in deep learning, and developing ML models better adapted to the characteristics of microbiome data will improve the performance and generalizability of models and enable their usage in the clinic.