Abstract

Abstract

Abstract

Abstract
Machine Learning has been overtaken by a growing necessity to explain and understand decisions made by trained models as regulation and consumer awareness have increased. Alongside understanding the inner workings of a model comes the task of verifying how adequately we can model a problem with the learned functions. Traditional global assessment functions lack the granularity required to understand local differences in performance in different regions of the feature space, where the model can have problems adapting. Residual Analysis adds a layer of model understanding by interpreting prediction residuals in an exploratory manner. However, this task can be unfeasible for high-dimensionality datasets through hypotheses and visualizations alone. In this work, we use weak interpretable learners to identify regions of high prediction error in the feature space. We achieve this by examining the absolute residuals of predictions made by trained regressors. This methodology retains the interpretability of the identified regions. It allows practitioners to have tools to formulate hypotheses surrounding model failure on particular regions for future model tunning, data collection, or data augmentation on critical cohorts of data. We present a way of including information on different levels of model uncertainty in the feature space through the use of locally fitted Model Agnostic Prediction Intervals (MAPIE) in the identified regions, comparing this approach with other common forms of conformal predictions which do not take into account findings from weak segment identification, by assessing local and global coverage of the prediction intervals. To demonstrate the practical application of our approach, we present a real-world industry use case in the context of inbound retention call-centre operations for a Telecom Provider to determine optimal pairing between a customer and an available assistant through the prediction of contracted revenue. © 2023, The Author(s), under exclusive license to Springer Nature Switzerland AG.

Abstract
This paper presents the main contributions of the VCMI Team to the ImageCLEFmedical Caption 2023 task. We addressed both the concept detection and caption prediction tasks. Regarding concept detection, our team employed different approaches to assign concepts to medical images: multi-label classification, adversarial training, autoregressive modelling, image retrieval, and concept retrieval. We also developed three model ensembles merging the results of some of the proposed methods. Our best submission obtained an F1-score of 0.4998, ranking 3rd among nine teams. Regarding the caption prediction task, our team explored two main approaches based on image retrieval and language generation. The language generation approaches, based on a vision model as the encoder and a language model as the decoder, yielded the best results, allowing us to rank 5th among thirteen teams, with a BERTScore of 0.6147. © 2023 Copyright for this paper by its authors.

Abstract
Emerging wearable technology able to monitor electrocardiogram (ECG) continuously for long periods of time without disrupting the patient's daily life represents a great opportunity to improve suboptimal current diagnosis of paroxysmal atrial fibrillation (AF). However, its integration into clinical practice is still limited because the acquired ECG recording is often strongly contaminated by transient noise, thus leading to numerous false alarms of AF and requiring manual interpretation of extensive amounts of ECG data. To improve this situation, automated selection of ECG segments with sufficient quality for precise diagnosis has been widely proposed, and numerous algorithms for such ECG quality assessment can be found. Although most have reported successful performance on ECG signals acquired from healthy subjects, only a recent algorithm based on a well-known pre-trained convolutional neural network (CNN), such as AlexNet, has maintained a similar efficiency in the context of paroxysmal AF. Hence, having in mind the latest major advances in the development of neural networks, the main goal of this work was to compare the most recent pre-trained CNN models in terms of classification performance between high- and low-quality ECG excerpts and computational time. In global values, all reported a similar classification performance, which was significantly superior than the one provided by previous methods based on combining hand-crafted ECG features with conventional machine learning classifiers. Nonetheless, shallow networks (such as AlexNet) trended to detect better high-quality ECG excerpts and deep CNN models to identify better noisy ECG segments. The networks with a moderate depth of about 20 layers presented the best balanced performance on both groups of ECG excerpts. Indeed, GoogLeNet (with a depth of 22 layers) obtained very close values of sensitivity and specificity about 87%. It also maintained a misclassification rate of AF episodes similar to AlexNet and an acceptable computation time, thus constituting the best alternative for quality assessment of wearable, long-term ECG recordings acquired from patients with paroxysmal AF.