TY - Generic T1 - Semantic Enrichment of XAI Explanations for Healthcare T2 - 24th International Conference on Artificial Intelligence Y1 - 2022 A1 - Corbucci, Luca A1 - Anna Monreale A1 - Cecilia Panigutti A1 - Michela Natilli A1 - Smiraglio, Simona A1 - Dino Pedreschi AB - Explaining black-box models decisions is crucial to increase doctors' trust in AI-based clinical decision support systems. However, current eXplainable Artificial Intelligence (XAI) techniques usually provide explanations that are not easily understandable by experts outside of AI. Furthermore, most of the them produce explanations that consider only the input features of the algorithm. However, broader information about the clinical context of a patient is usually available even if not processed by the AI-based clinical decision support system for its decision. Enriching the explanations with relevant clinical information concerning the health status of a patient would increase the ability of human experts to assess the reliability of the AI decision. Therefore, in this paper we present a methodology that aims to enable clinical reasoning by semantically enriching AI explanations. Starting from a medical AI explanation based only on the input features provided to the algorithm, our methodology leverages medical ontologies and NLP embedding techniques to link relevant information present in the patient's clinical notes to the original explanation. We validate our methodology with two experiments involving a human expert. Our results highlight promising performance in correctly identifying relevant information about the diseases of the patients, in particular about the associated morphology. This suggests that the presented methodology could be a first step toward developing a natural language explanation of AI decision support systems. JF - 24th International Conference on Artificial Intelligence ER - TY - CONF T1 - Doctor XAI: an ontology-based approach to black-box sequential data classification explanations T2 - Proceedings of the 2020 Conference on Fairness, Accountability, and Transparency Y1 - 2020 A1 - Cecilia Panigutti A1 - Perotti, Alan A1 - Dino Pedreschi AB - Several recent advancements in Machine Learning involve black-box models: algorithms that do not provide human-understandable explanations in support of their decisions. This limitation hampers the fairness, accountability and transparency of these models; the field of eXplainable Artificial Intelligence (XAI) tries to solve this problem providing human-understandable explanations for black-box models. However, healthcare datasets (and the related learning tasks) often present peculiar features, such as sequential data, multi-label predictions, and links to structured background knowledge. In this paper, we introduce Doctor XAI, a model-agnostic explainability technique able to deal with multi-labeled, sequential, ontology-linked data. We focus on explaining Doctor AI, a multilabel classifier which takes as input the clinical history of a patient in order to predict the next visit. Furthermore, we show how exploiting the temporal dimension in the data and the domain knowledge encoded in the medical ontology improves the quality of the mined explanations. JF - Proceedings of the 2020 Conference on Fairness, Accountability, and Transparency UR - https://dl.acm.org/doi/pdf/10.1145/3351095.3372855?download=true ER - TY - CONF T1 - Explaining multi-label black-box classifiers for health applications T2 - International Workshop on Health Intelligence Y1 - 2019 A1 - Cecilia Panigutti A1 - Riccardo Guidotti A1 - Anna Monreale A1 - Dino Pedreschi AB - Today the state-of-the-art performance in classification is achieved by the so-called “black boxes”, i.e. decision-making systems whose internal logic is obscure. Such models could revolutionize the health-care system, however their deployment in real-world diagnosis decision support systems is subject to several risks and limitations due to the lack of transparency. The typical classification problem in health-care requires a multi-label approach since the possible labels are not mutually exclusive, e.g. diagnoses. We propose MARLENA, a model-agnostic method which explains multi-label black box decisions. MARLENA explains an individual decision in three steps. First, it generates a synthetic neighborhood around the instance to be explained using a strategy suitable for multi-label decisions. It then learns a decision tree on such neighborhood and finally derives from it a decision rule that explains the black box decision. Our experiments show that MARLENA performs well in terms of mimicking the black box behavior while gaining at the same time a notable amount of interpretability through compact decision rules, i.e. rules with limited length. JF - International Workshop on Health Intelligence PB - Springer UR - https://link.springer.com/chapter/10.1007/978-3-030-24409-5_9 ER - TY - JOUR T1 - Assessing the use of mobile phone data to describe recurrent mobility patterns in spatial epidemic models JF - Royal Society open science Y1 - 2017 A1 - Cecilia Panigutti A1 - Tizzoni, Michele A1 - Bajardi, Paolo A1 - Smoreda, Zbigniew A1 - Colizza, Vittoria AB - The recent availability of large-scale call detail record data has substantially improved our ability of quantifying human travel patterns with broad applications in epidemiology. Notwithstanding a number of successful case studies, previous works have shown that using different mobility data sources, such as mobile phone data or census surveys, to parametrize infectious disease models can generate divergent outcomes. Thus, it remains unclear to what extent epidemic modelling results may vary when using different proxies for human movements. Here, we systematically compare 658 000 simulated outbreaks generated with a spatially structured epidemic model based on two different human mobility networks: a commuting network of France extracted from mobile phone data and another extracted from a census survey. We compare epidemic patterns originating from all the 329 possible outbreak seed locations and identify the structural network properties of the seeding nodes that best predict spatial and temporal epidemic patterns to be alike. We find that similarity of simulated epidemics is significantly correlated to connectivity, traffic and population size of the seeding nodes, suggesting that the adequacy of mobile phone data for infectious disease models becomes higher when epidemics spread between highly connected and heavily populated locations, such as large urban areas. VL - 4 ER -