1 March 2023 in International Journal of Reconfigurable and Embedded Systems (IJRES)
Intensive care unit deals with data that are dynamic in nature like real time measurement of health condition to laboratory test data that are continuously changes accordingly with time. Artificial intelligence (AI’s) potential ability to perform complex pattern analyses using large volumes of data. Generated pattern discovers the new symptoms of the disease in the Intensive care units (ICUs), helps the doctors to prescribe the new drug discovery which is helpful to intelligent use. Currently research work has been focused in the ICU making more efficient clinical workflow by generation of high-risk patterns from improved high volumes of data. Emerging area of AI in the ICU includes mortality prediction, uses of powerful sensors, new drug discovery, prediction of length of stay and legal role in uses of drugs for severity of disease. This review focuses latest application of AI drugs and other relevant issues for the ICU.
8 July 2026 in Research Square
Fine-tuning transformer-based handwritten text recognition (HTR) models on medieval manuscripts is challenging because these models are pre-trained on modern text and must adapt to a very different visual domain. This paper studies how three controllable finetuning choices (contrast normalization, data augmentation, and layer freezing) affect recognition accuracy when adapting TrOCR to small historical datasets. We run controlled experiments on a 13th-century Italian manuscript (I-CT 91 “Cortonese”) and replicate the same experimental grid on the public READ-16 benchmark as robustness evidence. On Cortonese, our best configuration achieves 8.03% character error rate (CER). Statistical comparisons across 13 configurations show that freezing up to three encoder layers or six decoder layers does not significantly harm accuracy, while deeper freezing becomes progressively detrimental. Removing contrast normalization (CLAHE) yields 7.84% CER, comparable to a domain-specialized baseline, suggesting strong optimization can reduce reliance on image preprocessing. Cross-dataset validation on READ-16 shows that decoder freezing thresholds transfer more robustly than encoder thresholds, and combined freezing strategies require dataset-specific re-validation. Finally, we use Grad-CAM gradient attributions and decoder cross-attention maps to diagnose error patterns and failure modes revealed by the ablations. Source code is available at https://github.com/LaudareProject/TrOCR-analysis