In an era characterized by rapid digital transformation, the convergence of technology and healthcare is redefining both clinical practices and patient experiences. The second edition of the Innovative Interfaces in Digital Healthcare (INI-DH) workshop seeks to gather innovators, healthcare practitioners, designers, and researchers to investigate the diverse applications of visual interfaces within the digital health domain.

Abstract In conventional machine learning (ML), a fundamental assumption is that the training and test sets share identical feature distributions, a reasonable premise drawn from the same dataset. However, real-world scenarios often defy this assumption, as data may originate from diverse sources, causing disparities between training and test data distributions. This leads to a domain shift, where variations emerge between the source and target domains. This study delves into human action recognition (HAR) models within an unconstrained, real-world setting, scrutinizing the impact of input data variations related to contextual information and video encoding. The objective is to highlight the intricacies of model performance and interpretability in this context. Additionally, the study explores the domain adaptability of HAR models, specifically focusing on their potential for re-identifying individuals within uncontrolled environments. The experiments involve seven pre-trained backbone models and introduce a novel analytical approach by linking domain-related (HAR) and domain-unrelated (re-identification (re-ID)) tasks. Two key analyses addressing contextual information and encoding strategies reveal that maintaining the same encoding approach during training results in high task correlation while incorporating richer contextual information enhances performance. A notable outcome of this study is the comprehensive evaluation of a novel transformer-based architecture driven by a HAR backbone, which achieves a robust re-ID performance superior to state-of-the-art (SOTA). However, it faces challenges when other encoding schemes are applied, highlighting the role of the HAR classifier in performance variations.

In this poster, we present a multimodal conversational interface, named Muxi, aiming to simplify meeting management in a collaborative work environment. The interaction is performed in two modalities: (i) a vocal conversation with a chatbot avatar and (ii) an interactive board where supporting information is provided.

In an era defined by digital transformation, the intersection of technology and healthcare is reshaping the landscape of medical practice and patient experience. The Innovative Interfaces in Digital Healthcare (INI-DH) workshop is at the forefront of this revolutionary evolution, addressing the key role of visual interfaces in healthcare. The workshop aims to bring together innovators, healthcare professionals, designers, and researchers to explore the multiple applications of visual interfaces in digital healthcare. Nine papers were accepted.

The application of artificial intelligence (AI) in medical imaging has significantly improved diagnostic accuracy. However, the reliance on black-box models remains a barrier to its widespread adoption in clinical settings. This article compares the main techniques of explainable AI (XAI), such as Grad-CAM, LIME, and SHAP, evaluating their effectiveness in interpreting deep learning models used in medical imaging. Two case studies are analyzed, comparing the three methods and highlighting their strengths and weaknesses. The results of this analysis show that Grad-CAM provides intuitive visualizations; LIME offers excellent flexibility in application; and SHAP delivers complete and accurate explanations, despite its high computational load.

Trust is a fundamental component of effective therapeutic relationships, significantly influencing patient engagement and treatment outcomes in mental health care. This paper presents a preliminary study aimed at enhancing trust through the co-creation of virtual therapeutic environments using generative artificial intelligence (AI). We propose a multimodal AI model, integrated into a virtual reality (VR) platform developed in Unity, which generates three-dimensional (3D) objects from textual descriptions. This approach allows patients to actively participate in shaping their therapeutic environment, fostering a collaborative atmosphere that enhances trust between patients and therapists. The methodology is structured into four phases, combining non-immersive and immersive experiences to co-create personalized therapeutic spaces and 3D objects symbolizing emotional or psychological states. Preliminary results demonstrate the system’s potential in improving the therapeutic process through the real-time creation of virtual objects that reflect patient needs, with high-quality mesh generation and semantic coherence. This work offers new possibilities for patient-centered care in mental health services, suggesting that virtual co-creation can improve therapeutic efficacy by promoting trust and emotional engagement.

Diagnosing rare diseases is difficult due to the complexity of the conditions, limited data, and a lack of specialized expertise. With over 10,000 rare diseases affecting more than 350 million people globally, diagnosis is often delayed or inaccurate, partly because traditional methods rely on fragmented and decentralized data. In this contribution, we highlight an issue similar to the curse of dimensionality that impacts the artificial intelligence training process, where too many features may lead to training failure. We named this issue the curse of heterogeneity: the need for massive interactions that slow down or lead to fail diagnosis process. Then, the contribution examines the challenges hidden behind rare disease diagnoses and discusses how SAI can improve it by combining AI-driven data analysis with human expertise. To do this, we use two real use-case scenarios. Finally, we discussed how SAI could optimize diagnosis processes and better use platforms like Orphanet, RareCare, and OMIM, which centralize rare disease data. The contribution aims to show how SAI offers a transformative approach to rare disease diagnosis by improving data integration, expert collaboration, and patient outcomes to expand the knowledge network as much as possible.

Technological evolution has enabled the development of new artificial intelligence (AI) models with generative capabilities. Among them, one of the most discussed is the virtual agent ChatGPT. This chatbot may occasionally produce fake information, as also declared by the producer OpenAI. Such a model may provide very useful support in several tasks, ranging from text summarization to programming. The research community has marginally investigated the impact that fake information created by AI models has on the users’ perceptions and on their belief in AI. We analyzed the impact of the fake information produced by AI on user perceptions, specifically trust and satisfaction, by performing a user study on ChatGPT. An additional issue is assessing whether the early or late knowledge of the possibility of the tool generating fake information has a different impact on the users’ perceptions. We conducted an experiment, involving 62 university students, a category of users who may employ tools such as ChatGPT extensively. The experiment consisted of a guided interaction with ChatGPT. Some of the participants experienced the failure of the chatbot, while a control group only received correct and reliable answers. We collected participants’ perceptions of trust, satisfaction, and usability, together with the net promoter score (NPS). The results demonstrated a statistically significant difference in trust and satisfaction between the users who early experienced fake information production compared to those who discovered ChatGPT’s faulty behaviors later during the interaction. Also, there is no statistically significant difference among the users who received the late fake information and the control group (no fake information). Usability and the NPS also resulted higher when the fake news was detected in the late interaction. When users are aware of the fake information generated by ChatGPT their trust and satisfaction decrease, especially when they impact on this at the early stage of use of the chatbot. Nevertheless, the perception of trust and satisfaction still remains high, as some of the users are still enthusiastic; others consider a more conscious use of the tool in terms of support to be verified. A useful strategy could be to favor a critical use of ChatGPT, letting young people to verify the provided information. This should be a new way to perform learning activities.

BACKGROUND: Melanoma is one of the deadliest tumors in the world. Early detection is critical for first-line therapy in this tumor pathology and it remains challenging due to the need for histological analysis to ensure correctness in diagnosis. Therefore, multiple computer-aided diagnosis (CAD) systems working on melanoma images were proposed to mitigate the need of a biopsy. However, although the high global accuracy is declared in literature results, the CAD systems for the health fields must focus on the lowest false negative rate (FNR) possible to qualify as a diagnosis support system. The final goal must be to avoid classification type 2 errors to prevent life-threatening situations. Another goal could be to create an easy-to-use system for both physicians and patients. RESULTS: To achieve the minimization of type 2 error, we performed a wide exploratory analysis of the principal convolutional neural network (CNN) architectures published for the multiple image classification problem; we adapted these networks to the melanoma clinical image binary classification problem (MCIBCP). We collected and analyzed performance data to identify the best CNN architecture, in terms of FNR, usable for solving the MCIBCP problem. Then, to provide a starting point for an easy-to-use CAD system, we used a clinical image dataset (MED-NODE) because clinical images are easier to access: they can be taken by a smartphone or other hand-size devices. Despite the lower resolution than dermoscopic images, the results in the literature would suggest that it would be possible to achieve high classification performance by using clinical images. In this work, we used MED-NODE, which consists of 170 clinical images (70 images of melanoma and 100 images of naevi). We optimized the following CNNs for the MCIBCP problem: Alexnet, DenseNet, GoogleNet Inception V3, GoogleNet, MobileNet, ShuffleNet, SqueezeNet, and VGG16. CONCLUSIONS: The results suggest that a CNN built on the VGG or AlexNet structure can ensure the lowest FNR (0.07) and (0.13), respectively. In both cases, discrete global performance is ensured: 73% (accuracy), 82% (sensitivity) and 59% (specificity) for VGG; 89% (accuracy), 87% (sensitivity) and 90% (specificity) for AlexNet.

Abstract Emotions recognition is widely applied for many tasks in different fields, from human-computer and human-robot interaction to learning platforms. Also, it can be used as an intrinsic approach for face recognition tasks, in which an expression-independent face classifier is developed. Most approaches face the problem by designing deeper and deeper neural networks that consider an expression as a still image or, in some cases, a sequence of consecutive frames depicting the temporal component of the expression. However, these suffer the training phase’s computational burden, which can take hours or days to be completed. In this work, a Web Shaped Model is proposed, which consists of a geometrical approach for extracting discriminant features from a face, depicting the characteristics of an expression. The model does not need to be trained since it is applied on a face and centred on the nose tip, resulting in image size and face size independence. Experiments on publicly available datasets show that this approach reaches comparable and even better results than those obtained applying DNN-based approaches.

Smartphones contain personal and private data to be protected, such as everyday communications or bank accounts. Several biometric techniques have been developed to unlock smartphones, among which ear biometrics represents a natural and promising opportunity even though the ear can be used in other biometric and multi-biometric applications. A problem in generalizing research results to real-world applications is that the available ear datasets present different characteristics and some bias. This paper stems from a study about the effect of mixing multiple datasets during the training of an ear recognition system. The main contribution is the evaluation of a robust pipeline that learns to combine data from different sources and highlights the importance of pre-training encoders on auxiliary tasks. The reported experiments exploit eight diverse training datasets to demonstrate the generalization capabilities of the proposed approach. Performance evaluation includes testing with collections not seen during training and assessing zero-shot cross-dataset transfer. The results confirm that mixing different sources provides an insightful perspective on the datasets and competitive results with some existing benchmarks.

Deep learning techniques have led to a vast improve-ment in various fields of computer vision, mainly using large-scale labelled datasets. Obtaining a large dataset of medical images for diagnostics is still an open challenge; the most significant obstacles are data imbalances and privacy issues related to sensitive patient information. Furthermore, the limited size of datasets for the training of a neural network can affect the performance of supervised learning and cause model overfitting problems. For this reason, data augmentation techniques are used to expand existing datasets. Generative Adversarial Networks (GANs) represent an innovative solution for acquiring additional information from a dataset since they can generate synthetic samples indistinguishable from real sample images. This work explores the use of GAN networks on Digital Breast Tomosynthe-sis (DBT) images, which is, in our knowledge, a completely new approach in this domain. In particular, we apply an optimization approach to the learning process of GAN networks based on evolutionary techniques.

Abstract According to the Wall Street Journal, one billion surveillance cameras will be deployed around the world by 2021. This amount of information can be hardly managed by humans. Using a Inflated 3D ConvNet as backbone, this paper introduces a novel automatic violence detection approach that outperforms state-of-the-art existing proposals. Most of those proposals consider a pre-processing step to only focus on some regions of interest in the scene, i.e., those actually containing a human subject. In this regard, this paper also reports the results of an extensive analysis on whether and how the context can affect or not the adopted classifier performance. The experiments show that context-free footage yields substantial deterioration of the classifier performance (2% to 5%) on publicly available datasets. However, they also demonstrate that performance stabilizes in context-free settings, no matter the level of context restriction applied. Finally, a cross-dataset experiment investigates the generalizability of results obtained in a single-collection experiment (same dataset used for training and testing) to cross-collection settings (different datasets used for training and testing).

In this work, we combine Curriculum Learning with Deep Reinforcement Learning to learn without any prior domain knowledge, an end-to-end competitive driving policy for the CARLA autonomous driving simulator. To our knowledge, we are the first to provide consistent results of our driving policy on all towns available in CARLA. Our approach divides the reinforcement learning phase into multiple stages of increasing difficulty, such that our agent is guided towards learning an increasingly better driving policy. The agent architecture comprises various neural networks that complements the main convolutional backbone, represented by a ShuffleNet V2. Further contributions are given by (i) the proposal of a novel value decomposition scheme for learning the value function in a stable way and (ii) an ad-hoc function for normalizing the growth in size of the gradients. We show both quantitative and qualitative results of the learned driving policy.

Abstract Head pose estimation represents an important computer vision technique in different contexts where image acquisition cannot be controlled by an operator, making face recognition of unknown subjects more accurate and efficient. In this work, starting from partitioned iterated function systems to identify the pose, different regression models are adopted to predict the angular value errors (yaw, pitch and roll axes, respectively). This method combines the fractal image compression characteristics, such as self-similar structures in order to identify similar head rotation, with regression analysis prediction. The experimental evaluation is performed on widely used benchmark datasets, i.e., Biwi and AFLW2000, and the results are compared with many existing state-of-the-art methods, demonstrating the robustness of the proposed fusion approach and excellent performance.

Head pose estimation is a sensitive topic in video surveillance/smart ambient scenarios since head rotations can hide/distort discriminative features of the face. Face recognition would often tackle the problem of video frames where subjects appear in poses making it quite impossible. In this respect, the selection of the frames with the best face orientation can allow triggering recognition only on these, therefore decreasing the possibility of errors. This paper proposes a novel approach to head pose estimation for smart cities and video surveillance scenarios, aiming at this goal. The method relies on a cascade of two models: the first one predicts the positions of 68 well-known face landmarks; the second one applies a web-shaped model over the detected landmarks, to associate each of them to a specific face sector. The method can work on detected faces at a reasonable distance and with a resolution that is supported by several present devices. Results of experiments executed over some classical pose estimation benchmarks, namely Point '04, Biwi, and AFLW datasets show good performance in terms of both pose estimation and computing time. Further results refer to noisy images that are typical of the addressed settings. Finally, examples demonstrate the selection of the best frames from videos captured in video surveillance conditions.