Artificial Neural Network-Assisted Ultrasound Image Capture on Personal Computers

Abstract

Traditional probes consist of 40 to 60 crystals each attached to a pin attached to a specialized cable of 
minimum 2 to 3 m. Each crystal has piezoelectric properties. To capture images by personal computers 
to improve accessibility and reduce the cost of having ultrasonic image with special reference to 
obstetrics and gynecology emergency settings, a probe was designed consisting of three Doppler 
transducers (each with 4 pins hence generating 4 signals, altogether 12 signal) that by an analog 
switching with micro (1000 HZ per second) can generate 12000 signals per second changing the scan 
line form linear into a plane. The signals are translated into WAV sound format file that can be displayed 
by a Windows-based program of a personal computer. The pattern produced is created by the sound of 
blood flow in an organ. This vascular pattern was matched with traditional sonography of the organ. By 
training the network, the resolution of images can be improved further based on the formula: output = 
0.77*target + 38. The probe can capture images 1/6 resolution of a traditional probe, deeper penetration 
(19 cm depth), 1/34 price and weight of a traditional ultrasound equipment