Portail HAL de l'Université Picardie Jules Verne
Not a member yet
32425 research outputs found
Sort by
Compte rendu de : Frédéric Worms, Avec Bergson, Paris, PUF, collection « Philosophie française contemporaine », 2024, 524 p.
No full textInternational audienc
L’enfer du tour d’échelle en droit des biens : rétrospectives et perspectives
No full textInternational audienc
Correction: Histamine H3 Receptor as a target for alcohol use disorder: challenging the predictability of animal models for clinical translation in drug development
No full textInternational audienc
Perceived Normal and Pathological Aging? A Cross-cultural Comparison Between French and Congolese
No full textInternational audienceDespite well-established cultural nuances regarding the perception of aging, there is a paucity of research specifically examining how different cultures distinguish between normal aging processes and pathological conditions in older adults. The present study aimed to offer insights into the socio-cultural influences on this distinction. It addressed this gap by comparing the perception of 516 French and 210 Congolese individuals. A specially designed 55-item questionnaire depicting various situations involving older individuals was administered in paper format. Participants assessed each situation on a Likert-type scale as indicative of either normal or pathological aging. The situations depicted cognitive, behavioral, and emotional changes commonly associated with neurocognitive disorders as an illustration of pathological aging. Data were analyzed using descriptive statistics, ANOVA, and linear regression modeling. In line with our hypothesis, French participants rated more severely the situations compared to Congolese participants. Moreover, French participants performed better in differentiating between situations illustrating normal or pathological aging, while Congolese individuals considered all situations as reflecting normal aging. This research showed that perception of normal or pathological behavior in aging is not universally shared across cultures. It also revealed that the general population lacks scientific knowledge on normal and pathological aging, outlining a need for improvement regarding public education. Differences between populations also suggest that public education should be specifically tailored and contextualized to improve knowledge on aging. We encourage further studies on African populations from a neuropsychological perspective for better representativeness of the human species and to facilitate access to unbiased scientific knowledge
Comparison of oral water ingestion and intravenous fluid infusion on fluid responsiveness in healthy volunteers, a prospective, randomized trial
No full textInternational audienc
From the ground to the roof: 3D microclimate modelling in a tropical forest and its consequences for epiphytic bryophytes
No full textInternational audienc
Guest editorial – ICDCIT 2024 & 2025
No full textPart of special issue : Selected Papers from the 20th and 21th International Conferences on Distributed Computing and Intelligent Technology (ICDCIT 2024 & 2025), January 2024 and 2025, Kalinga Institute of Industrial Technology (KIIT), Bhubaneswar, Indi
Multimodal Deep Learning Model for BCLC Staging of Hepatocellular Carcinoma: Integration of Radiological Imaging and Clinical Data
No full textInternational audienceBackground&Aims: Hepatocellular carcinoma (HCC) staging according to the BCLC (Barcelona Clinic Liver Cancer) system requires integration of tumor burden, liver function, and performance status. Despite promising AI applications in HCC diagnosis, no validated multimodal models exist for automated BCLC staging through integrated analysis of radiological imaging and clinical text reports. We developed a multimodal deep learning model for BCLC stage classification.Methods: We conducted a retrospective study of 137 consecutive HCC patients at the University of Palermo (June 2019-May 2023), analysing 247 radiological examinations (192 CT, 55 MRI) with corresponding clinical reports. Patients were classified into six categories: five BCLC stages (0, A, B, C, D) and complete response (CR) status. We developed a multimodal neural architecture combining a convolutional neural network (CNN) for radiological image feature extraction and BioBERT for natural language processing of clinical reports. Text data were pre-processed using regular expressions to identify key patterns regarding metastases, nodule characteristics, and portal vein involvement. The model was trained using a two-phase strategy with an 80:20 train/test split.Results: The cohort included 247 radiological assessments with class distribution: CR 37.2%, stage 0 4.5%, stage A 14.2%, stage B 7.3%, stage C 23.9%, and stage D 4.5%. The multimodal model achieved 66.9% overall accuracy on the test set. Confusion matrix analysis revealed excellent predictive accuracy for the CR class, directly correlated with its higher dataset representation. Performance metrics demonstrated significant class-specific variability: the model showed high discrimination for CR cases but substantial difficulty distinguishing between adjacent BCLC stages. The confusion matrix demonstrated a tendency to misclassify cases between adjacent BCLC stages, suggesting the model captures a general disease severity gradient but lacks fine-grained discriminative capacity for clinically similar presentations.Conclusions: This study demonstrates feasibility of a multimodal AI approach for HCC BCLC staging integrating radiological imaging and clinical text data. While preliminary results show suboptimal performance compared to binary diagnostic tasks, they reflect the inherent complexity of multi-class BCLC staging. The model serves as a proof-of-concept for clinical decision support and staging standardization
