Abstract
[No abstract available]

Abstract
The U-Net is one of the most fundamental architectural advancements in the deep learning era. It is a crucial tool for image segmentation, especially for biomedical modalities. The research community seems to interpret the effectiveness of neural architectural search (such as the nn-U-Net) as evidence that architectural enhancements proposed since its debut are mostly unnecessary. We argue that there are still network-in-network primitives that can be leveraged to further enhance its performance, focusing on the squeeze-and-excitation (SE) pathway specifically in this paper. Specifically, we study its use of global descriptors, since it should be at odds with the spatial resolution required for dense-prediction tasks. It is theorized in the literature that performance is probably gained from some implicit ability of the learned excitations to filter supposedly uninformative channels during training. We explain this almost unreasonable success through an analysis of the empirical estimates of the excitation covariance matrix. Our analysis also directly contradicts the above conjecture - the most effective SE approach actually displayed the less extreme filtering behaviour, weighing all channels much closer to the mean (0.5). Our experiments are conducted in three diverse, staple biomedical modalities: dermoscopy, colonoscopy, and ultrasound.

Abstract

Abstract

Abstract
The main challenges hindering the adoption of deep learning-based systems in clinical settings are the scarcity of annotated data and the lack of interpretability and trust in these systems. Concept Bottleneck Models (CBMs) offer inherent interpretability by constraining the final disease prediction on a set of human-understandable concepts. However, this inherent interpretability comes at the cost of greater annotation burden. Additionally, adding new concepts requires retraining the entire system. In this work, we introduce a novel two-step methodology that addresses both of these challenges. By simulating the two stages of a CBM, we utilize a pretrained Vision Language Model (VLM) to automatically predict clinical concepts, and an off-the-shelf Large Language Model (LLM) to generate disease diagnoses grounded on the predicted concepts. Furthermore, our approach supports test-time human intervention, enabling corrections to predicted concepts, which improves final diagnoses and enhances transparency in decision-making. We validate our approach on three skin lesion datasets, demonstrating that it outperforms traditional CBMs and state-of-the-art explainable methods, all without requiring any training and utilizing only a few annotated examples. The code is available at https://github.com/CristianoPatricio/2step-concept-based-skin-diagnosis.

Abstract
Rising competitiveness and client requirements make effective use of high volume and complexity real-time industrial data crucial for faster decision-making. However, this potential is hindered by a lack of smart, user-friendly analytic tools for all collaborators. Despite the proliferation of Machine Learning (ML) tools for data scientists, non-experts struggle with converting data into actionable insights and identifying profitable data science projects. A smart tool is thus proposed, allowing non-experts to perform preliminary data evaluations through profiled analysis pathways that execute predefined sets of Exploratory Data Analysis (EDA) methods and ML operations. Further assisting users, the tool solely relies on metadata attributes and textual descriptions of datasets enhanced by interaction with a Large Language Model (LLM). This paper examines profile selection stages, replacing traditional ML methods with Conformal Prediction (CP) techniques. CP identifies multiple potential prospects with statistical confidence and recognizes when correct predictions are impossible. Trials with task-labeled metadata files (derived from publicly available datasets) showed that while classic ML methods had about 80% efficiency, CP techniques improved the selection process, keeping profiling errors below 0.06 with 99% confidence. This approach enables the correct identification (with statistical confidence) of appropriate analysis profiles for data science problems, thus paving the way for more efficient data analysis tools in industrial settings, accessible to users of all skill levels. © 2024 The Authors. Published by Elsevier B.V.