Ultrasonic data processing method and device, ultrasonic equipment and storage medium

Abstract

The invention discloses an ultrasonic data processing method and device, ultrasonic equipment and a medium. The method comprises the steps of acquiring and loading all dynamic link libraries; whereinthe dynamic link library is a plug-in corresponding to a reasoning algorithm model; after a service processing request is received, determining a dynamic link library required by the service processing request, and loading a target reasoning algorithm model corresponding to the dynamic link library; and controlling the target reasoning algorithm model to process the ultrasonic data corresponding to the service processing request to obtain an ultrasonic data processing result. According to the invention, the reasoning algorithm model is expressed in the form of a dynamic link library; all dynamic link libraries are loaded after a program is started, only the dynamic link libraries are loaded, only a small memory is occupied, a video memory is not occupied, and the target reasoning algorithmmodel corresponding to the dynamic link library is loaded only when a service processing request needs to be responded, so that the combination of a deep learning network and ultrasonic diagnosis isrealized; and ultrasonic acquisition data does not need to be copied to a high-performance computer, so that the data processing efficiency is improved.

Description

technical field [0001] The present application relates to the technical field of deep learning, and more specifically, relates to an ultrasonic data processing method and device, an ultrasonic device, and a computer-readable storage medium. Background technique [0002] Ultrasound equipment combined with deep learning reasoning network when diagnosing images can provide analysis results based on big data learning experience, and can provide auxiliary diagnostic value for clinical medicine, pathological analysis and confirmation. At the same time, using the pattern recognition technology of the deep learning network, it can automatically classify and divide the ultrasonic images within a certain threshold range, which greatly reduces the difficulty of the doctor's operation and saves diagnosis time. When the ultrasound image data is processed using a deep learning network, the offline mode is generally used, that is, the ultrasound acquisition data is copied to a high-perform...

AI Technical Summary

Problems solved by technology

For example, the more data the hardware can infer in a unit of time, the more accurate the results will be; if the inference data in a unit of time is less, data processing can only be performed by skipping frames, resulting in a decrease in the accuracy of the results

Currently, the graphics card devices that support deep learning are mainly NVIDIA graphics cards, whose performance is 50% higher than that of the GTX9 series, but the price is twice as high, and the cooling cost will also increase

Because the ultrasound equipment is different from the PC machine, the internal structure also includes front-end board, probe slot, touch screen cable, ECG card and other hardware, so there is limited cooling space for the graphics card. It is also impossible to achieve the same large power supply as a PC machine

That is to say, general ultrasound equipment will consider factors such as heat dissipation, hardware cost, space occupied by the machine, and cost, and often choose hardware equipment with relatively medium hardware indicators such as small size and computing performance, which cannot be fully matched in hardware configuration. Ultra-high-performance, high-heating deep learning workstations have equal performance, making it difficult to combine deep learning with ultrasound diagnosis

Images