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.

Abstract
Cardiovascular diseases are one of the main causes of death in the world. The predominant form of cardiovascular disease is coronary artery disease. Coronary artery calcium scanning is a non-contrast computed tomography exam that is considered the most reliable predictor of coronary events. Deep learning models have been developed for the segmentation of coronary artery calcium but the results have limited interpretability due to the black-box nature of these models. This work proposes an image retrieval pipeline based on a supervised contrastive framework that is capable of enhancing this interpretability by providing similar visual examples of coronary calcifications. In the COCA dataset, it is shown that this retrieval presents a label precision of 0.944 +/- 0.230 regarding artery labels of retrieved images, with moderate similarity in terms of calcification area and Agatston score. It is also shown that the retrieval can be used to correct a deep CAC segmentation model by passing predictions from a segmentation model through the retrieval system, improving robustness and explainability.

Abstract
Neuroblastoma (NB) is the most common extracranial tumor in pediatric cases. The MYCN oncogene amplification (MNA) is knowingly correlated with a poor prognosis, making detecting this biomarker crucial for treatment selection and survival prediction. The current clinical protocol for MNA detection includes invasive procedures, such as biopsy. The proposed work aims to develop non-invasive techniques for predicting MNA in patients with diagnosed NB, using AI-based models and Computerized Tomography (CT) scans. Machine learning methods that use the imaging features extracted from the tumor on the CT slices were developed and compared with deep learning (DL) models. Additionally, agnostic explainable methods for imaging were applied to create explanations about the relevant information used by the DL models in the prediction. The results show a better performance for the DL approach, which achieved an AUC of 0.94 ± 0.04. The similarity in the explanations produced by the models trained with different data splits showed that feature extraction remains somewhat invariant to shifts in training data, which is relevant given the small amount of data available. Learning models were shown to have predictive potential that, with further improvements, can be integrated into predictive, explainable, and, thus, trustworthy systems to aid clinicians in the decision-making process. © 2025 IEEE.