UniCas for Medicine and Healthcare Marco Cantone1,† , Svonko Galasso1,2,3,† , Gabriele Lozupone1,† , Emanuele Nardone1,† , Cesare Davide Pace1,† , Ciro Russo1,† , Alessandro Bria1,† , Tiziana D’Alessandro1,*,† , Claudio De Stefano1,† , Francesco Fontanella1,† , Claudio Marrocco1,† , Mario Molinara1,† and Alessandra Scotto di Freca1,† 1 Dept. of Electrical and Information Engineering (DIEI), University of Cassino and Southern Lazio, Via G. Di Biasio 43, 03043 Cassino (FR), Italy 2 Dept. of Research and Development, LUNEX, 50 Avenue du Parc des Sports, 4671 Differdange, Luxembourg 3 Luxembourg Health & Sport Sciences Research Institute A.s.b.l., 50 Avenue du Parc des Sports, 4671 Differdange, Luxembourg Abstract With over twenty years of experience, our research group, affiliated with the Artificial Intelligence and Data Analysis Laboratory (AIDA), which belongs to the University of Cassino and Southern Lazio (UniCas), has been deeply engaged in artificial intelligence. The specific focus on Machine Learning, Pattern Recognition, and Deep Learning has evolved theoretically, with the development of specialized skills in model optimization, and practically, through application to real-world problems, particularly in the healthcare domain. In particular, attention is paid to designing and implementing Computer-Aided Diagnosis systems to support the prevention, diagnosis, and monitoring of Neurodegenerative diseases, Specific Learning Disorders, breast cancer, diabetic retinopathy, and movement-related disorders. Different data is utilized to reach the objectives of the AIDA Lab, and many approaches are implemented. Handwriting analysis is exploited to support the diagnosis of neurodegenerative diseases and specific learning disorders and to monitor them over time. Handwriting analysis encompasses two distinct approaches: examining dynamic features and scrutinizing handwriting sample images. This comprehensive approach allows a more thorough understanding of the individual’s writing characteristics. Additionally, in Neurodegenerative Diseases, advancements include the utilization of 3D image analysis of MRI scans to aid in the detection of Alzheimer’s disease, further enhancing diagnostic capabilities in this field. Mammograms are used for breast cancer prevention and diagnosis, while retinal images are used for diabetic retinopathy detection, particularly focusing on detecting small lesions. Detecting small lesions is a crucial step in diagnosis, as is identifying microcalcifications in digital mammograms and microaneurysms in digital fundus images. To address this challenge, we propose a novel architecture called GravityNet. Another field of study in the research conducted by the AIDA Lab is movement analysis, focusing on gait analysis, enabling precise evaluations in real-world settings and potential applications in Parkinson’s disease assessment. To this end, Machine and advanced Deep Learning techniques are employed, such as deep cascades of boosting classifiers and Deep Convolutional and Attentional Neural Networks. Keywords Neurodegenerative Diseases, Handwriting, 3D Image Analysis, Breast Cancer, Small Lesion, Movement Analysis, Retinopathy 1. Neurodegenerative Diseases Ital-IA 2024: 4th National Conference on Artificial Intelligence, orga- and Brain Disorders nized by CINI, May 29-30, 2024, Naples, Italy * Corresponding author. The prevalence of neurodegenerative diseases (NDs) has † These authors contributed equally. been steadily increasing in recent years, underscoring $ marco.cantone@unicas.it (M. Cantone); a concerning trend. Another aspect underscoring the svonko.galasso@unicas.it (S. Galasso); gabriele.lozupone@unicas.it importance of research in this field is that NDs currently (G. Lozupone); emanuele.nardone@unicas.it (E. Nardone); lack a cure. They can cause cognitive impairments man- cesaredavide.pace@unicas.it (C. D. Pace); ciro.russo@unicas.it (C. Russo); a.bria@unicas.it (A. Bria); tiziana.dalessandro@unicas.it ifesting as difficulties in memory, language, thinking, (T. D’Alessandro); destefano@unicas.it (C. De Stefano); judgment, and motor skills. Individuals exhibiting a com- fontanella@unicas.it (F. Fontanella); c.marrocco@unicas.it bination of these symptoms face a significantly height- (C. Marrocco); m.molinara@unicas.it (M. Molinara); ened risk of developing dementia and, in more severe a.scotto@unicas.it (A. Scotto di Freca) cases, Alzheimer’s Disease (AD) or Parkinson’s Disease  0000-0002-6225-6680 (M. Cantone); 0000-0001-5163-4458 (S. Galasso); 0009-0001-7377-9866 (G. Lozupone); (PD). Given the progressive nature of these conditions, 0009-0005-8718-5435 (E. Nardone); 0009-0006-5678-1610 early detection becomes crucial to initiate therapies to (C. D. Pace); 0009-0002-8751-8605 (C. Russo); 0000-0003-0840-7350 mitigate their effects. (A. Bria); 0009-0005-4340-7227 (T. D’Alessandro); Concerning other brain-related disorders, the prevalence 0000-0002-7654-6849 (C. De Stefano); 0000-0002-3242-0179 of specific learning disorders (SLD) presents a significant (F. Fontanella); 0000-0003-0840-7350 (C. Marrocco); 0000-0002-6144-0654 (M. Molinara); 0000-0002-9642-3555 (A. Scotto and persistent challenge in educational and developmen- di Freca) tal psychology. These disorders encompass a range of © 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 conditions affecting an individual’s ability to acquire and air movements up to a maximum height of 3𝑐𝑚 from apply foundational academic skills, such as reading, writ- its surface. After the acquisition step, many images of ing, and mathematics, despite adequate intelligence and the handwritten traces were collected or generated. Pa- opportunities for learning. Unlike NDs, which primar- per sheets were scanned to generate the offline image ily affect cognitive functioning in adulthood, specific dataset [2], while the raw data, expressed as coordinates, learning disorders often manifest during childhood and were used to generate synthetic images by interpolating can persist into adolescence and adulthood if left un- consecutive points. In this way, we generated the binary treated. Given the chronic nature of specific learning dataset [3] and the RGB one [4], which encoded dynamic disorders and their potential to impact multiple domains information in the three color channels. Figure 1 illus- of functioning, early identification and intervention are trates the various images examined and produced within paramount. this research path. In particular, they refer to the Clock Drawing Test. 1.1. Handwriting Handwriting analysis has emerged as a valuable tool in the early detection of NDs like AD. This method’s significance lies in its ability to provide insights into var- ious cognitive, memory, and motor functions integral to Figure 1: Samples of different kinds of Handwriting images writing. The complexity of writing, which encompasses acquired and generated, specifically focusing on the Clock language proficiency, executive skills, and attention span, Drawing Test. can reveal changes in linguistic complexity, word choice, and syntactic structures. Such alterations are pivotal for early AD diagnosis as they reflect the underlying cog- These datasets were used to feed a system that ex- nitive changes. Additionally, the link between writing ploited the ability of Convolutional Neural Networks and memory offers a window into the patient’s ability (CNNs) to extract features automatically. Various experi- to associate names with concepts, identify temporal rela- ments were conducted using different task configurations tionships, and recall previously seen words, highlighting and datasets. Despite deploying numerous experimental declines in episodic memory. Moreover, the analysis of settings, they all shared a common baseline architec- fine motor control through handwriting, such as changes ture, facilitating comparative analysis and yielding in- in pressure, speed, and stroke patterns, can indicate earlysightful results. The baseline experimental architecture declines in motor functions. These aspects of handwrit- comprises four key steps: data acquisition, feature ex- ing analysis serve as critical markers for the early iden- traction/engineering, Machine Learning (ML) and Deep tification and intervention of AD, aiming to delay its Learning (DL) classification, and combining rule applica- progression and improve patient outcomes. tion. The outcomes of this research revealed interesting trends. Across various tasks, CNN-extracted deep fea- tures consistently outperformed other approaches, em- 1.1.1. Task Images Analysis phasizing the advantage of DL. Experimentation with In the context of AD, alterations in handwriting are of- offline images suggests their potential in diagnosing NDs, ten observable due to the cognitive and motor changes underscoring the importance of considering shape details associated with the condition. In the 2000s, researchers in distinguishing patients from healthy individuals, espe- began to recognize the potential of handwriting analysis cially in medical applications where sensitivity is crucial. as a non-invasive and accessible tool for early detection. Finally, this research highlighted that some handwriting In 2018, researchers from the AIDA LAb started a metic- tasks were more significant in supporting AD diagnosis, ulous data acquisition campaign by administering an and combining them improved the final prediction. experimental protocol [1] of 25 handwriting tasks. The research included a total of 174 participants, with 89 1.1.2. Dynamic Analysis patients diagnosed with AD and 85 healthy control sub- Our research also focused on handwriting analysis by jects. Each participant was invited to perform the exper- capturing raw data through digital devices, which track imental protocol by using the WACOM Bamboo Folio intricate handwriting details like pressure, tilt, and mo- graphic tablet, enabling participants to write on standard tion. This data was meticulously organized and analyzed A4 white paper sheets using a pen that appears typical. to extract critical dynamic and static features, such as This pen not only produced ink traces on the sheet but velocity, acceleration, and stroke pressure, providing a also generated digital information recorded by the tablet rich dataset for examination. Before the ML algorithms, in the form of spatial coordinates and pressure for each we conducted thorough preprocessing to normalize and point (𝑥, 𝑦, and 𝑧). The tablet additionally captured in- clean the data, followed by feature selection processes [5] to highlight the most indicative characteristics of the handwriting. This preparation phase also included hy- perparameter tuning to refine our ML models for optimal performance. Our methodology incorporated two dis- tinct ML approaches: a Statistical Approach, analyzing overall handwriting patterns to detect deviations indica- tive of diseases, and a Stroke Approach, focusing on the detailed analysis of individual handwriting strokes. This dual perspective enabled us to capture both the broad characteristics and the fine-grained patterns in handwrit- ing that may signal the early stages of conditions like Alzheimer’s disease. By leveraging digital technology for data acquisition and applying a nuanced ML analy- sis, our study aims to enhance the early diagnosis and Original 3D MRI Explainable 3D MRI understanding of NDs, offering the potential for timely intervention and improved patient outcomes. Figure 2: From a 3D MRI image, the network can highlight the AD-related brain regions. 1.2. Specific Learning Disorder Handwriting and drawing are essential for preschool and 1.3. 3D Images Analysis school-age children’s cognitive and motor development. Studies indicate that a significant portion of the school 3D imaging is a widely used technique in diagnostic pro- day, ranging from 30-60%, is dedicated to fine motor activ- cedures for brain disorders. Magnetic resonance imaging ities, primarily handwriting. Graphomotor tasks, crucial (MRI) provides a non-invasive means to observe and an- for academic success and personal expression, account alyze in vivo pathological changes in the brain related for about 40% of a primary school student’s activities. to Brain Disorders, facilitating the study of disease evo- In this age group, 20% of children are at risk of expe- lution. MRI analysis is significant in AD diagnosis since riencing graphomotor difficulties, with 27% displaying it is identified by structural and functional changes in subpar graphomotor abilities. It is essential to note the dynamically changing morphological patterns, which role of Specific Learning Disorders (SLDs), such as dys- are appropriately captured with high-resolution MRI. It graphia, which are closely associated with graphomotor is also noteworthy that brain atrophy, a distinctive AD challenges. Early detection and intervention are crucial symptom, can be identified through MRI. This form of in addressing these disorders. Our research aims to de- atrophy serves as a reliable marker of the disease. It is velop innovative ML solutions that promptly identify and indicative of its progression, as well as being associated address handwriting issues early in a child’s academic with tau deposition and neuropsychological impairments, journey. Timely intervention is crucial, as the window essential factors in the clinical manifestation of AD. for effective rehabilitation narrows over time. ML is es- In recent years, CNNs and Vision Transformers have sential for identifying subtle yet vital changes in the transformed neuroimaging data analysis for AD, moving graphomotor cycle that experienced specialists may miss. beyond the traditional ML that focuses on approaches The study started with creating an experimental protocol that rely on handcrafted features and classifiers. Recent to reveal the fundamental features of handwriting. The works have proven promising in diagnosing and predict- handwriting was digitized using graphic tablets, generat- ing Alzheimer’s; however, at the same time, presenting ing a comprehensive dataset for training ML models to different limitations mainly linked to the use of 3D CNN accurately differentiate between various learning impair- networks. Training these networks requires many sam- ments. The investigation will be longitudinal, spanning ples, often unavailable in the medical field. One way four years of data collection. This will increase the depth to mitigate this problem is to use 3D volume slicing ap- of the dataset and allow for a detailed exploration of proaches. The volume is divided into 2D slices by slicing the evolution of graphomotor skills and associated chal- from one of the three anatomical body planes. This solu- lenges. The monitoring will enable the study to identify tion partially solves the problem but involves the loss of graphomotor problem trends and indicators, enhancing learning global patterns of the entire volume. ML algorithms’ capacity to classify handwriting issues Another key aspect of DL models in the medical field and lead to accurate early interventions. is the interpretability of their results. It is crucial for clin- ical applications and remains a significant challenge that hinders the deployment of DL-aided diagnostic systems in real-world scenarios. Our research aims to develop models that compensate for single calcification detection in mammograms involv- for current limitations. We are developing approaches ing a specialized convolution layer operating in the fre- that allow us to be effective with few training samples quency domain as the first layer of a CNN [7]. This layer and are explainable at the same time. Radiologists typ- automatically learns a bank of Difference of Gaussians ically examine 3D images as a series of 2D images and (DoG) band-pass filters parameterized by their associated base their diagnosis by focusing on specific brain parts pairs of standard deviations inversely proportional to the from different views. Inspired by this, we have developed filter’s bandwidth. Additionally, we explore the potential a diagnosis and XAI approach that allows us to diagnose of transformer models in backbone-head architecture for AD and, at the same time, highlight the areas predomi- lesion detection [8]. nantly affected by the disease. From a series of 2D slices Amalgamating traditional image processing tech- extracted from the volume, the network can reconstruct niques with ML and DL methodologies has significantly a three-dimensional saliency map highlighting the areas advanced cancer detection in breast imaging. Ongoing affected by AD with voxel-level precision (see Figure 2). exploration of innovative approaches, such as frequency- Preliminary results show effectiveness in highlighting filter-integrated CNNs and transformer-based models, areas commonly associated with the disease, suggesting offers promise for further enhancing detection accuracy. that these approaches are promising for diagnosing and predicting brain disorders. 2.2. Small lesion detection There has been significant interest in utilizing DL tech- 2. Medical Image Analysis niques for medical image analysis in recent years. These methods, employing advanced neural network architec- 2.1. Breast cancer detection tures, have greatly improved the ability to extract valu- Breast cancer ranks as the most prevalent form of cancer able information from complex medical images. This ad- among women and stands as the second leading cause vancement holds promise for more accurate diagnostics of death. Detecting breast cancer at an early stage is and tailored healthcare solutions. While current object crucial for increasing survival rates, prompting the im- detection models have shown impressive success, the in- plementation of mammography screening programs in creasing resolution of medical images presents a unique numerous countries. Mammograms can reveal various challenge, requiring greater attention to detail and iden- abnormalities, including calcifications, masses, architec- tifying smaller objects within the images. Small lesions tural distortions, and focal asymmetry. can be critical indicators of various medical conditions, Computer-aided diagnosis (CAD) systems have been making their accurate detection crucial for diagnosis and devised to aid radiologists in analyzing mammographic treatment planning. images for diagnostic and screening purposes. Incor- To address this, a novel DL model, called GravityNet porating ML and DL techniques has led to significant [9], is proposed to enhance the detection of small objects, progress in numerous tasks, such as image classification, employing innovative anchoring techniques known as lesion detection, and segmentation. CNNs stand as the gravity points, that appear to be ’attracted’ to lesions. de facto standard in many image analysis applications, The architecture comprises a backbone network and two demonstrating the ability to detect microcalcifications subnets dedicated to regression and classification tasks. and other types of lesions accurately. Our research re- This proposed approach notably enhances detection ac- volves around utilizing CNNs and transformer models curacy, leading to earlier diagnoses and improved patient to address various challenges in mammogram analysis. outcomes. The results underscore the significance of this Transformer-based models, such as the Vision Trans- innovative approach, demonstrating its broad implica- former, promise to capture a broader context in breast tions across various clinical applications. imaging tasks. Ongoing research on transformer-based For instance, GravityNet is applied to identify calcifi- models in mammography indicates their efficacy in tasks cations in digital mammograms, which are crucial for the like multi-view mammogram classification. early detection of breast cancer. Furthermore, its effec- One of our research interests involves the classification tiveness in detecting microaneurysms in retinal images of entire mammograms as malignant or normal, assess- could significantly aid the early diagnosis and manage- ing their performance based on the type of lesion present ment of diabetic retinopathy, a leading cause of blindness and at different input resolutions [6]. Mammographic worldwide. images exhibit lesions with diverse characteristics in size, In addition, GravityNet shows promise in cytology shape, texture, and sparsity, which may align better with imaging, particularly in the analysis of whole slide im- local convolutional paradigms or self-attention capable of ages to detect cell nuclei, which is crucial for cervical can- capturing long-range features. We also propose a method cer identification. Accurate identification of cell nuclei in cytology images helps pathologists identify morphology and abnormal cell patterns indicative of cervical cancer. This capability has significant potential to improve the efficiency and accuracy of cervical cancer screening pro- grams, enabling earlier diagnosis and intervention to reduce the morbidity and mortality associated with this prevalent cancer. Figure 3: From 2D input video to marker-less joint and seg- Finally, GravityNet extends to the identification of ment angles assessment large vessel occlusions (LVOs), a crucial task in stroke diagnosis and treatment planning. LVOs are obstructions in the brain’s main arteries, and their early detection is need for markers attached to the patient’s skin, streamlin- essential to determine appropriate treatment strategies ing the experimental workflow and reducing preparation to minimize neurological damage and improve patient time. Moreover, these systems allow to perform GAn outcomes. Leveraging its advanced DL algorithms, Grav- even outside gait laboratories. This makes them useful ityNet can accurately detect and localize LVOs in volu- for biomechanical evaluations in sports settings and as- metric space within computed tomographic angiography sessing ambulatory functions in patients with neuromo- (CTA) scans. This capability allows healthcare profession- tor impairments during routine activities. This approach als to quickly identify patients at risk of a severe stroke combines movement and GAn with clinical precision and and promptly initiate interventions such as thrombec- practical applicability, which allows for patient care and tomy or thrombolysis, which can significantly reduce athletic performance enhancement in real-world settings. disability and mortality rates. MMA systems bridge the gap between rigorous labora- tory analysis and everyday assessment, representing a 3. Movement Analysis significant stride in precision medicine and rehabilitative care. Figure 3 shows the practical application of an inno- Gait Analysis (GAn) is an objective assessment of a per- vative MMA framework developed to estimate joint and son’s walking abilities and is an essential part of mo- segment angles from 2D videos and images. It showcases tor assessment. It helps health professionals make in- real-time data acquisition and analysis during a tread- formed clinical decisions and develop targeted rehabil- mill gait assessment, proving the successful integration itation strategies to improve gait functions. In clinical of technology and biomechanics in modern therapeutic practice, GAn is performed by healthcare professionals and athletic environments. By delivering precise and via standardized questionnaires, functional tests, and vi- dependable joint and segment dynamics measurements, sual observation of the patient’s walking pattern. The such a solution can significantly augment human move- identification of gait irregularities is based on the sub- ment analysis, potentially elevating health outcomes and jective quantification of spatio-temporal parameters and athletic performance. detailed kinematic and kinetic evaluations. Moreover, in the context of Parkinson’s Disease (PD), Various advanced technologies have completely a common neurodegenerative disorder with progressive changed the way of analyzing motion objectively. Op- loss of dopaminergic and other sub-cortical neurons, GAn tical motion capture systems are the most accurate and is an essential tool for the diagnosis and progression of precise for joint kinematics assessment, making them the the disease, which is characterized by motor symptoms "gold standard" within laboratory settings. Such tech- such as tremors, rigidity, bradykinesia (slowness of move- nologies enable the accurate acquisition of motion data ment), and postural instability. Through the utilization using reflective markers placed at strategic joint locations of GAn, clinicians can better understand how PD affects and multiple cameras that track them. However, they an individual’s mobility in daily life, enabling the identi- are mainly restricted to specialized gait laboratories and fication of specific gait patterns associated with PD pro- research settings due to their limitations, which include gression. To achieve this aim, ML algorithms are used to high costs, technical expertise demands, and extensive classify diseases based on spatiotemporal and kinematic setup durations, hampering widespread clinical adoption. gait features extracted from MMA. An essential aspect ML approaches are increasingly used in GAn, developing of employing ML in medical diagnostics is explainability, from integrating computational intelligence with biome- particularly when understanding the impact of different chanics to offer robust methods for analyzing complex features on the model’s predictions. Understanding what movement patterns [10]. CNN are increasingly utilized features have a major impact on the outcome is essential for human pose estimation within marker-less motion in the medical field. The use of MMA systems presents a analysis (MMA) systems, significantly augmenting their significant advancement in the management of PD, en- precision and reliability. In contrast to marker-based abling continuous and real-time monitoring of patients’ motion analysis systems, MMA systems eliminate the movements. This continual observation facilitates the development of highly personalized rehabilitation and raphy classification: an experimental study, Sensors therapy programs that can be adjusted daily to meet the 23 (2023) 1229. evolving needs of each patient. [7] M. Cantone, C. Marrocco, F. Tortorella, A. Bria, Learnable dog convolutional filters for microcalcifi- cation detection, Artificial Intelligence in Medicine Acknowledgments 143 (2023) 102629. [8] A. S. Betancourt Tarifa, C. Marrocco, M. Molinara, 1. Part of these researches were funded by Italian F. Tortorella, A. Bria, Transformer-based mass de- Ministry of University, MIUR program “Diparti- tection in digital mammograms, Journal of Ambient menti di Eccellenza 2018-2022” Law 232/216 and Intelligence and Humanized Computing 14 (2023) by D.M. 351/2022 “Dottorati innovativi per la pub- 2723–2737. blica amministrazione”. [9] C. Russo, A. Bria, C. Marrocco, Gravitynet for 2. Project ECS 0000024 “Ecosistema end-to-end small lesion detection, Artificial Intelli- dell’innovazione - Rome Technopole” fi- gence in Medicine (2024) 102842. doi:10.1016/j. nanced by EU in NextGenerationEU plan artmed.2024.102842. through MUR Decree n. 1051 23.06.2022 PNRR [10] S. Galasso, R. Baptista, M. Molinara, S. Pizzocaro, Missione 4 Componente 2 Investimento 1.5 - R. S. Calabrò, A. M. De Nunzio, Predicting physical CUP H33C22000420001. activity levels from kinematic gait data using 3. Supported by Luxembourg National Research machine learning techniques, Engineering Appli- Fund (FNR) (MEMENTO – Machine lEarning- cations of Artificial Intelligence 123 (2023) 106487. based Marker-lEss gait analysis system for clini- URL: https://www.sciencedirect.com/science/ cal assessment of humaN moTiOn [16749075]) article/pii/S0952197623006711. doi:https://doi. org/10.1016/j.engappai.2023.106487. References [1] N. Cilia, C. De Stefano, F. Fontanella, A. Scotto di Freca, An experimental protocol to support cog- nitive impairment diagnosis by using handwriting analysis, Procedia Computer Science 141 (2018) 466 – 471. [2] N. Dalia Cilia, T. D’Alessandro, C. De Stefano, F. Fontanella, Offline handwriting image analy- sis to predict alzheimer’s disease via deep learn- ing, in: 2022 26th International Conference on Pattern Recognition (ICPR), 2022, pp. 2807–2813. doi:10.1109/ICPR56361.2022.9956359. [3] N. D. Cilia, T. D’Alessandro, C. De Stefano, F. Fontanella, M. Molinara, From online handwrit- ing to synthetic images for alzheimer’s disease de- tection using a deep transfer learning approach, IEEE Journal of Biomedical and Health Informat- ics 25 (2021) 4243–4254. doi:10.1109/JBHI.2021. 3101982. [4] N. D. Cilia, T. D’Alessandro, C. De Stefano, F. Fontanella, Deep transfer learning algorithms applied to synthetic drawing images as a tool for supporting alzheimer’s disease prediction, Machine Vision and Applications 33 (2022) 49. [5] T. D’Alessandro, C. De Stefano, F. Fontanella, E. Nar- done, A. Scotto di Freca, Feature evaluation in handwriting analysis for alzheimer’s disease using bayesian network, in: International Graphonomics Conference, Springer, 2023, pp. 122–135. [6] M. Cantone, C. Marrocco, F. Tortorella, A. Bria, Con- volutional networks and transformers for mammog-