Integrating digital twins and deep learning for medical image analysis in the era of COVID-19 - Medical Xpress

Digital twins are virtual representations of devices and processes that seizure the carnal properties of the situation and operational algorithms/techniques successful the discourse of aesculapian devices and technologies. Digital twins whitethorn let wellness attraction organizations to find methods of improving aesculapian processes, enhancing diligent experience, lowering operating expenses, and extending the worth of care.

During the contiguous COVID-19 pandemic, assorted medical devices, specified arsenic X-rays and CT scan machines and processes, are perpetually being utilized to cod and analyse medical images. When collecting and processing an extended measurement of information successful the signifier of images, machines and processes sometimes endure from strategy failures, creating captious issues for hospitals and patients.

To code this, a survey published successful Virtual Reality & Intelligent Hardware introduces a digital-twin-based astute wellness attraction strategy integrated with aesculapian devices to cod accusation regarding the existent wellness condition, configuration, and attraction past of the device/machine/system. Furthermore, aesculapian images, that is, X-rays, are analyzed by utilizing a deep-learning exemplary to observe the corruption of COVID-19.

The designed strategy is based connected the cascade recurrent convolution neural network (RCNN) architecture. In this architecture, the detector stages are deeper and much sequentially selective against tiny and adjacent mendacious positives. This architecture is simply a multi-stage hold of the RCNN exemplary and sequentially trained utilizing the output of 1 signifier for grooming the other. At each stage, the bounding boxes are adjusted to find a suitable worth of the nearest mendacious positives during the grooming of the antithetic stages. In this manner, the statement of detectors is adjusted to summation the intersection implicit union, overcoming the occupation of overfitting. This survey trains the exemplary by utilizing X-ray images arsenic the exemplary was antecedently trained connected different dataset.

The developed strategy achieves bully accuracy during the detection signifier of COVID-19. The experimental outcomes uncover the ratio of the detection architecture, which yields a mean mean precision complaint of 0.94.

More information: Imran Ahmed et al, Integrating Digital Twins and Deep Learning for Medical Image Analysis successful the epoch of COVID-19, Virtual Reality & Intelligent Hardware (2022). DOI: 10.1016/j.vrih.2022.03.002

Provided by Beijing Zhongke Journal Publising Co.