Towards AI-driven Next Generation Personalized Healthcare and Well-being Fatih Aksu1 , Alessandro Bria2 , Alice Natalina Caragliano3 , Camillo Maria Caruso3 , Wenting Chen4 , Ermanno Cordelli3,* , Omar Coser3,5 , Arianna Francesconi3 , Leonardo Furia3 , Valerio Guarrasi3 , Giulio Iannello3 , Clemente Lauretti5 , Guido Manni3,5 , Giustino Marino3 , Domenico Paolo3 , Filippo Ruffini3 , Linlin Shen6 , Rosa Sicilia3 , Paolo Soda3,7 , Christian Tamantini5 , Matteo Tortora3 , Zhuoru Wu6 and Loredana Zollo5 1 Department of Biomedical Sciences, Humanitas University, Milan, Italy 2 Department of Electrical and Information Engineering, University of Cassino and Southern Latium, Cassino, Italy 3 Unit of Computer Systems and Bioinformatics, Department of Engineering, University Campus Bio-Medico of Rome, Italy 4 City University of Hong Kong 5 Unit of Advanced Robotics and Human-Centered Technologies, Department of Engineering, University Campus Bio-Medico of Rome, Italy 6 Shenzhen University 7 Department of Diagnostics and Intervention, Radiation Physics, Biomedical Engineering, Umeå University, Sweden Abstract In the last few years Artificial Intelligence (AI) is emerging as a game changer in many areas of society and, in particular, its integration in medicine heralds a transformative approach towards personalized healthcare and well-being, promising significant improvements in diagnostic precision, therapeutic outcomes, and patient care. Our research explores the cutting- edge realms of multimodal AI, resilient AI, and healthcare robotics, aiming to harness the synergy of diverse data modalities and advanced computational models to redefine healthcare paradigms. This multidisciplinary effort seeks to bridge technology and clinical practice, advancing AI-driven next generation personalized healthcare and well-being. Keywords Artificial Intelligence, Multimodal Learning, Precision Medicine, Stress Detection, Resilient AI, Healthcare Robotics 1. Introduction 2. Multimodal AI enables precision Artificial Intelligence (AI) has proven itself as enabling medicine factor for triggering great transformations of society [1, The evolution of precision medicine marks a paradigm 2, 3, 4]. However on the verge of the fifth industrial shift from the traditional "one-size-fits-all" approach in revolution, there are several challenges that involve the healthcare towards tailored therapeutic strategies that consolidation of AI arrival in sectors as medicine and account for individual variability in genes, environment, people well-being. Indeed this paradigm shift towards AI- and lifestyle. In this context, leveraging the variety of pa- driven healthcare is not just a technological revolution; tient generated data (as images, clinical data, electronic it represents a comprehensive reimagining of medical health records etc.) can provide a significant boost to practices, enhancing the quality, efficiency, and accessi- unlocking a holistic view of the patient. Towards this bility of healthcare services. In this scenario our efforts end multimodal AI provides the ultimate tool [5, 6]: the are directed towards four research paths: (i) multimodal integration is not merely additive, it’s transformative, AI for precision medicine (section 2); (ii) multimodal AI enabling the extraction of insights that would remain to foster wellbeing (section 3); (iii) resilient AI (section 4); obscured under traditional, unimodal analysis. We are (iv) AI in robotics for healthcare (section 5). For each of currently studying the potential of multimodal AI for these routes we provide a brief description of the devel- precision medicine facing different challenges in differ- oped solutions, highlighting solved problems and open ent application domains: in the oncological domain we challenges. face challenges regarding data fusion and representa- tion, with two projects on Non-Small Cell Lung Can- cer (NSCLC) (sections 2.1 and 2.2); in augmenting the Ital-IA 2024: 4th National Conference on Artificial Intelligence, orga- nized by CINI, May 29-30, 2024, Naples, Italy diagnosis and prognosis of Alzheimer (section 2.3) we * Corresponding author. tackle the problem of imbalance in multimodal datasets; $ e.cordelli@unicampus.it (E. Cordelli) and in COVID-19 prognosis (section 2.4) we attempt to  0000-0001-6062-7575 (E. Cordelli) solve issues related to scarcity of large, labelled datasets © 2024 Copyright for this paper by its authors. Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0). CEUR ceur-ws.org Workshop ISSN 1613-0073 Proceedings with compatible tasks for training deep learning models slices crucial for predicting OS outcome. without leading to overfitting. 2.2. PICTURE 2.1. AIDA PICTURE stands for "Pathological response AI-driven AIDA stands for "explAinable multImodal Deep learning prediCTion after neoadjUvant theRapiEs in NSCLC". This for personAlized oncology". This project faces the chal- project is based on the central hypothesis that heteroge- lenge of advancing Multimodal Deep Learning (MDL), neous medical data (i.e. radiological images, histology studying how to learn shared representations between images, cytology and molecular data and EHRs) are con- different modalities, by investigating when to fuse the sistent with the pathological complete response (pCR), different modalities and how to embed in the training any so their combination using artificial intelligence (AI) can process able to learn more powerful data representations. provide accurate pCR prediction in NSCLC patients. In- All this is directed towards facing the association between deed, albeit treating locally advanced NSCLC surgically radiomic, pathomic and Electronic Health Records (EHRs) is the mainstay, it is important to prevent post-surgery in precision oncology to predict the patient outcomes in recurrence, and neoadjuvant therapy (NAT) has shown terms of progression free survival, overall survival, re- potential in enhancing overall survival rates and achiev- lapse time and response rate NSCLC, which represents ing a complete pathological response, that, if correctly the 85% of all lung cancer cases. To pursue these objec- evaluated before the treatment, can even avoid non nec- tives we started from learning unimodal representation essary surgical resections. of EHRs and medical imaging. PICTURE pursues three objectives: (i) pCr prediction As a prior contribution, EHRs are vital resources for through radiology imaging, histology, citology, molec- documenting patient clinical history and procedures, but ular data, EHRs, and their combination; (ii) leveraging are often challenging to process due to their unstructured multimodal deep learning to make the performance of nature. Natural Language Processing (NLP) tools, par- AI resilient and robust for pCR prediction signature; (iii) ticularly Named Entity Recognition (NER) with the use improving trust and transparency using explainable AI of Transformer-based models, have proven effective in models. PICTURE also has the exploratory aim of trans- extracting meaningful information from EHRs [7]. Trans- ferring trained models to predict pCR for patients under- formers excel at capturing contextual relationships be- going chemoimmunotherapy, tailoring treatments to the tween words and the still not thoroughly explored con- individual needs of patients. textual embedding they create can enhance the under- standing of the content itself. We propose the Hieararchi- 2.3. Facing imbalance in Alzheimer’s cal Embedding Attention for overall survivaL (HEAL), a methodology that leverages multi-class NER-driven rep- Disease diagnosis and prognosis resentations from EHRs by weighting them with atten- Alzheimer’s disease (AD) is a progressive neurodegen- tional mechanisms. The ability of emphasizing clinically erative condition with decline in cognitive function, relevant information within unstructured data, operating and because of the lack of a cure, its early detection is both at word and sentence levels, makes HEAL more in- paramount. Despite the recent progress in AI, challenges terpretable for medical applications. In a NSCLC Overall such as class imbalance, integration of multimodal data, Survival (OS) prediction case study, HEAL achieved an and robust generalization remain pervasive. In response average 𝐶 𝑡𝑑 -index of 0.639 and a low standard deviation to this we introduce a novel methodology that leverages of 0.014 over 5 runs, showing a statistically significant the strengths of ensemble learning while incorporating superiority with respect to manually extracted clinical advanced fusion techniques. For each of the 4 modali- features. ties of the tabular ADNI database, we train a series of Our second contribution, even if still at its prelimi- classifiers on varied class distributions followed by a late nary steps, grounds on the fact that deep learning (DL) fusion strategy that integrates the different modalities to approaches have demonstrated significant value in au- improve the results. tomatically learning potentially relevant patterns from Our framework is evaluated on two diagnostic tasks medical images, such as computed tomography (CT) [8]. (binary and ternary) and four binary prognostic tasks (at Hence, in this study we explore a novel methodology 12, 24, 36, and 48 months) and compared with 12 state- for predicting OS in NSCLC patients using only CT im- of-the-art imbalanced data algorithms, achieving 97.04% ages, aiming at a multitask architecture that encompasses g-mean on the binary diagnostic task and 90.81% g-mean prognostic factors like Progression-Free Survival (PFS) on the 48-month prognostic task. beyond predicting OS alone. The first steps in this direc- tion include producing a soft attention weighted feature map for each input slice and highlighting the relevant Physiological Environment DRL Agent 2.4. Multi-Dataset Multi-Task Learning • Preprocessing Oversampling Critic Actor for COVID-19 Prognosis • • • Standardization Class Balancing Data Augmentation Estimated discounted reward In COVID-19 context [9] in order to fight the scarcity of TD 0,1 0,1 0,1 large, labelled chest radiographic images (CXR) datasets, Growing Window error 𝑺𝒐𝒇𝒕𝒎𝒂𝒙 Physiological signals we introduce a novel multi-dataset multi-task (MDMT) #$%"&'( #$%"&'( 𝒎𝒂𝒙𝟏 𝑝" 𝒎𝒂𝒙𝟐 𝑝" Reward Function training framework, by integrating correlated datasets False #$%"&'( Δ = 𝒎𝒂𝒙𝟏 𝑝" R = −α𝑡! −𝒎𝒂𝒙𝟐 𝑝" #$%"&'( >τ from disparate sources and assessing severity score to True classify prognostic severity groups [10, 11, 12], instead of 𝒔𝒐𝒇𝒕𝒎𝒂𝒙 +𝒑𝒕 𝑖𝑓 𝑐𝑜𝑟𝑟𝑒𝑐𝑡 𝑝𝑟𝑒𝑑𝑖𝑐𝑡𝑖𝑜𝑛 * 𝒔𝒐𝒇𝒕𝒎𝒂𝒙 −𝒑𝒕 𝑖𝑓 𝑖𝑛𝑐𝑜𝑟𝑟𝑒𝑐𝑡 𝑝𝑟𝑒𝑑𝑖𝑐𝑡𝑖𝑜𝑛 relying on datasets with multiple and correlated labelling Dynamic Psycho-Physical Observation Window schemes. As illustrated in figure 1 a deep CNN takes State the images as input and branches into task-specific fully Figure 2: Overview of the proposed method for early stress connected output networks, to end with a multi-task loss detection consisting of two main blocks: physiological envi- function incorporating an indicator function to exploit ronment and DRL agent. The first involves the pre-processing multi-dataset integration. of data and the description of the dynamic observation space. The second block incorporates a SAC-based DRL agent fed with data from the first block to control the system. Deep Reinforcement Learning (DRL); second, we are fur- ther expanding the multimodal view integrating infor- mation from video, audio and text with the physiolog- ical data. Robust and fast stress detection approaches Figure 1: Overview of the proposed Multi-Dataset Multi- can bring benefit in several contexts: from providing Task model architecture, composed by a shared backbone 𝑓 𝑠 a targeted and more personal assistance to patients, to and two task-specific fully connected network heads, 𝑓 𝜏1 and 𝑓 𝜏2 , for tasks 𝜏1 and 𝜏2 , respectively, producing outputs 𝑂𝜏1 ensuring safety for workers, for instance, Air Traffic Con- and 𝑂𝜏2 . trollers (ATC) that endure high levels of psychological pressure during their job impacting operational safety. Our first approach [13] employs a new DRL model to Proceeding with a 5 cross-validation and leave one identify stress indicators. We obtained this by leveraging center out training, we evaluated the method across 18 a dynamic time observation window that expands each different CNN backbone on prognosis classification task step of the learning process, asking the agent to choose and fine-tuning from BRIXIA dataset to AIforCOVID either to continue observing or to classify based on the dataset task. Best average performance with statistical information gathered until that point, trying to mini- robustness achieved: 68.6% accuracy, 66.6% F1-score and mize the amount of data required for decision-making. 68.5% g-mean for the 5 cross validation, and 65.7% accu- As depicted in the figure Figure 2, we adopted the Soft racy, 64.3% F1-score and 66.0% g-mean for the leave-one- Actor-Critic algorithm for its effectiveness in handling center-out validation strategy. Future directions include continuous action spaces. In a Leave-One-Subject-Out new domains and the integration of XAI [6]. approach with data augmentation on the Non-EEG pub- lic dataset we outperformed existing solutions, showing 3. Multimodal AI to foster the power of DRL for early stress detection. On top of this approach we are exploring the larger well-being multimodal asset of the Ulm-Trier Social Stress Test Stress, a response to physical and emotional demands, dataset (ULM-TSST, MuSe 2022 challenge), containing 41 is crucial in determining individuals’ well-being and if training, 14 validation and 14 test subjects, simulating a unmanaged can lead to conditions such as anxiety, depres- job interview scenario, with audio, video, text, and phys- sion and cardiovascular diseases. Also in this scenario iological data modalities, rated on arousal and valence multimodal AI offers a tool for proactive approach to stress parameters. The aim is to build a high performance health management, in order to provide real-time moni- architecture that leverages non invasive modalities for toring and interventions, thereby mitigating long-term stress detection that can be employed in work environ- health risks associated with chronic stress. ments. In both cases the scarcity of large datasets that We are targeting stress detection from two perspec- provide a quantitative measurement of stress is still the tives: first, we are focusing on maximising the stress level main challenge: we will try to face it considering the prediction accuracy within the shortest possible time, ex- construction of robust and specific acquisition protocols ploiting multimodal physiological time series data and to test the effectiveness of the developed approaches in real-world scenarios. 4. Resilient AI Due to the high stakes involved in healthcare decisions, the sensitivity of medical data, and the complexity of medical environments, AI systems should be designed to maintain their intended performance and integrity in the face of adversities, as data corruption, missing data, privacy leakage, or unexpected changes in their operating environment. This is the goal of Resilient AI, which is an aspect that cannot be left out when the aim Figure 3: Overall framework of the proposed method working is to integrate AI for augmenting the medical practice. with triplet networks. To meat this goal we are currently investigating three main aspects that fall under the Resilient AI umbrella: developing systems robust to missing data, the challenge Last but not least, the challenge related to patient pri- of limited extension datasets and how to protect sensitive vacy led us to explore Federated learning (FL). FL presents patients data. an innovative solution to the challenge of protecting sen- With respect to the missing data challenge [14], al- sitive patient data in artificial intelligence applications though a variety of strategies exist for addressing this in healthcare, in fact enabling the training of a shared problem in health datasets, to overcome the obstacle to global model with a central server while ensuring data select the most suitable one and their dependency on the privacy within local institutions. On this basis we intro- dataset’s specifics, we developed a Transformer-based duce a new token-based FL paradigm, revolutionizing the model [15] that applies masking to ignore the missing traditional approach with sequential or random passing data, thus eliminating the need of imputation and dele- of a token between clients during each epoch. This inno- tion techniques and focusing directly only on the avail- vative method allows only the token owner to send the able features through self-attention. Moreover, we in- weights to the server, which redistributes them directly troduced a novel feature-identifying form of positional to all models. By eliminating local training epochs and encoding to facilitate the integration of tabular data into allowing immediate transmission, this paradigm shift a Transformer framework. This method was validated streamlines the process by circulating a single model through an overall survival classification task, employing among clients and also mitigating the need for an initial clinical data from the CLARO [16] project and improving warm-up period, potentially paving the way for a de- the prediction accuracy. centralized system that reduces dependence on a central In order to address the problem of working with server and minimizes the number of parameters trans- datasets limited in the extension, particularly frequent mitted in each iteration. Results on the tabular part of in healthcare domain, Triplet networks, a subtype of the the AIforCOVID dataset [18] composed of 6 hospitals Siamese networks, emerge as a promising solution, com- show that the performance of the FL model does not de- prising three identical networks operating concurrently. viate from that of its equivalent trained on all datasets Throughout training of these three networks two inputs aggregated into a single pool. The next steps will focus belong to the same class, whereas the third belongs to a on integrating other modalities into the FL pipeline, such distinct class, with the final objective to develop a feature as CXR scans of the AIforCOVID dataset itself. space with two distinct clusters one per class by incor- porating inter-class diversities alongside intra-class simi- larities and providing scenarios with limited data with 5. AI for healthcare robotics more triplets compared to instances (Figure 3). In our study [17], using a private dataset of 86 CT scans, triplet The integration of robotics in healthcare settings exem- networks surpass the plain deep networks in accurately plifies another dimension of AI’s impact, automating rou- predicting the histological subtypes of NSCLC patients. tine tasks, assisting in surgeries with precision beyond Currently, we are broadening the scope of our research human capability, and providing rehabilitation support including PET images alongside CT scans and adopt- to patients. This not only enhances service delivery but ing a multimodal strategy for the same classification. also alleviates the workload on healthcare professionals, By integrating these complementary data we anticipate allowing them to focus more on patient-centered care. achieving a significant improvement and overcoming the In this scenario we pursue two aims: first, enhancing challenges posed by limited data scenarios. robotic surgery with real-time high precision localiza- friendly, efficient operation, and safety. The aim of our work is to recognize the terrain on which an exoskeleton is walking and its inclination. Among several state-of- the-art driven approaches, we achieved promising results using LSTM architectures with IMU data (0.94 of accu- racy in Leave-one-out cross-validation), and CNN-LSTM architectures with EMG data (0.75 of accuracy). The fu- sion of IMU and EMG data didn’t bring any significant improvement, as explanatory tests indicated that the best 20 contributing features belong to IMU. Next, by varying the number of sensors, and therefore features, we noticed Figure 4: The MVSLAM pipeline integrates depth estimation, that the best results are achieved by selecting the most pose estimation, and 3D reconstruction modules to generate relevant features, from one to three, according to SHAP a continuously updated 3D map of the surgical environment (on a 3 subjects validation set), leading to 0.85, 0.89 and from monocular endoscopic video frames. 0.93 of accuracy respectively. Lastly, we found that LSTM and CNN-LSTM are valid architectures for slope inclina- tion prediction (MAE of 1.95°) and stair height (MAE of tion; second, boosting lower-limb robotic rehabilitation 15.65 mm), without significant differences in employing optimizing the structural exoskeleton sensor configura- 3 or 4 sensors. tion. For the first objective we focus on the laparoscopy use case, as one of the preferred surgical methods. Despite re- Acknowledgments cent advancements in image acquisition it is still limited Fatih Aksu, Alice Natalina Caragliano, Camillo Maria to rely on 2D images view: misinterpreting anatomical Caruso, Omar Coser, Arianna Francesconi, Leonardo Fu- structures due to this limit is a common source of er- ria, Guido Manni, Giustino Marino, Domenico Paolo rors. In contrast, 3D imaging increases the accuracy of and Filippo Ruffini are Ph.D. students enrolled in instrument manipulation, leads to better outcomes in the National Ph.D. in Artificial Intelligence, course surgery, and shortens the learning for trainees. Even sev- on Health and life sciences, organized by Univer- eral research in surgical 3D imaging has been explored, sità Campus Bio-Medico di Roma. We acknowl- like camera-based tracking and mapping, Mosaicking, edge financial support from: i) PNRR MUR project Structure from Motion, and Shape from Template, they PE0000013-FAIR; ii) PRIN 2022 MUR 20228MZFAA- often rely on simplifications that can limit their effec- AIDA (CUP C53D23003620008); iii) PRIN PNRR 2022 tiveness. On this ground Simultaneous Localization and MUR P2022P3CXJ-PICTURE (CUP C53D23009280001); Mapping (SLAM) has shown promising results, as it aims iv) FCS MISE (CUP B89J23000580005); v) MAECI (grant to create a map of the environment while localizing the n. CN23GR09); vi) NRR MUR project PNC0000007 sensor position within it. Therefore we developed a ro- Fit4MedRob. This work was also partially supported bust deep learning SLAM pipeline to operate in real-time by the following companies: Eustema S.p.A. and ENAV across diverse surgical settings by providing an immer- S.p.A.. sive, interactive 3D environment (Figure 4), allowing for more precise and personalized interventions with the fu- ture possibility to be integrated with augmented reality References displays. For the second objective, we focus on the challenges in [1] V. Guarrasi, L. Tronchin, C. M. Caruso, A. Rofena, the field of lower limb robotics. It aims at supporting peo- G. Manni, F. Aksu, D. Paolo, G. Iannello, R. Sicilia, ple with lower limb disabilities by enhancing movement, E. Cordelli, et al., Building an ai-enabled metaverse mobility, and providing targeted exercise. Technologies for intelligent healthcare: opportunities and chal- as exoskeletons, prosthetics, and rehabilitation robots, lenges, in: Ital-IA 2023, Italia Intelligenza Artificiale are particularly helpful for those with neurological issues, Thematic Workshops, co-located with the 3rd CINI offering improved rehabilitation, independence, and tai- National Lab AIIS Conference on Artificial Intel- lored care. Effective use requires precise control settings ligence (Ital IA 2023), Pisa, Italy, May 29-30, 2023, to adapt walking patterns to different terrains. Chal- CEUR-WS, 2023, pp. 134–139. lenges in this field involve the extensive need for sensors [2] E. Cordelli, V. Guarrasi, G. Iannello, F. Ruffini, R. Si- for terrain detection and the complexity in processing cilia, P. Soda, L. Tronchin, Making ai trustworthy in sensor data. Simplifying sensor requirements to accu- multimodal and healthcare scenarios, Proceedings rately determine terrain and slope is critical for user- of the Ital-IA (2023). [3] J. P. Bharadiya, Artificial intelligence and the fu- [11] V. Guarrasi, N. C. D’Amico, R. Sicilia, E. Cordelli, ture of web 3.0: Opportunities and challenges ahead, P. Soda, Pareto optimization of deep networks American Journal of Computer Science and Tech- for covid-19 diagnosis from chest x-rays, Pattern nology 6 (2023) 91–96. Recognition 121 (2022) 108242. [4] V. Pereira, E. Hadjielias, M. Christofi, D. Vrontis, A [12] V. Guarrasi, P. Soda, Optimized fusion of cnns to systematic literature review on the impact of artifi- diagnose pulmonary diseases on chest x-rays, in: cial intelligence on workplace outcomes: A multi- International Conference on Image Analysis and process perspective, Human Resource Management Processing, Springer, 2022, pp. 197–209. Review 33 (2023) 100857. [13] L. Furia, et al., Exploring early stress detection from [5] V. Guarrasi, P. Soda, Multi-objective optimization multimodal time series with deep reinforcement determines when, which and how to fuse deep learning, in: 2023 IEEE International Conference networks: An application to predict covid-19 out- on Bioinformatics and Biomedicine (BIBM), IEEE, comes, Computers in Biology and Medicine 154 2023, pp. 1917–1920. (2023) 106625. [14] A. Rofena, V. Guarrasi, M. Sarli, C. L. Piccolo, [6] V. Guarrasi, L. Tronchin, D. Albano, E. Faiella, M. Sammarra, B. B. Zobel, P. Soda, A deep learn- D. Fazzini, D. Santucci, P. Soda, Multimodal ex- ing approach for virtual contrast enhancement in plainability via latent shift applied to covid-19 strat- contrast enhanced spectral mammography, arXiv ification, arXiv preprint arXiv:2212.14084 (2022). preprint arXiv:2308.00471 (2023). [7] D. Paolo, et al., Named entity recognition in ital- [15] C. M. Caruso, V. Guarrasi, S. Ramella, P. Soda, ian lung cancer clinical reports using transformers, A deep learning approach for overall survival in: 2023 IEEE International Conference on Bioin- analysis with missing values, arXiv preprint formatics and Biomedicine (BIBM), IEEE, 2023, pp. arXiv:2307.11465 (2023). 4101–4107. [16] CLARO - CoLlAborative multi-sources Radiopath- [8] M. Tortora, et al., Radiopathomics: multimodal omics approach for personalized Oncology in non- learning in non-small cell lung cancer for adaptive small cell lung cancer., http://www.cosbi-lab.it/ radiotherapy, IEEE Access (2023). claro/, 2020. Accessed: 2023-03-20. [9] G. Fiscon, F. Salvadore, V. Guarrasi, A. R. Garbuglia, [17] F. Aksu, et al., Early experiences on using triplet P. Paci, Assessing the impact of data-driven lim- networks for histological subtype classification in itations on tracing and forecasting the outbreak non-small cell lung cancer, in: 2023 IEEE 36th Inter- dynamics of covid-19, Computers in biology and national Symposium on Computer-Based Medical medicine 135 (2021) 104657. Systems (CBMS), IEEE, 2023, pp. 832–837. [10] V. Guarrasi, N. C. D’Amico, R. Sicilia, E. Cordelli, [18] P. Soda, N. C. D’Amico, J. Tessadori, G. Valbusa, P. Soda, A multi-expert system to detect covid-19 V. Guarrasi, C. Bortolotto, M. U. Akbar, R. Sicilia, cases in x-ray images, in: 2021 IEEE 34th Inter- E. Cordelli, D. Fazzini, et al., Aiforcovid: Predict- national Symposium on Computer-Based Medical ing the clinical outcomes in patients with covid-19 Systems (CBMS), IEEE, 2021, pp. 395–400. applying ai to chest-x-rays. an italian multicentre study, Medical image analysis 74 (2021) 102216.