Prostate scanning is commonly requested by physicians for male patients above a certain age for both diagnostic and screening purposes. The accurate determination of prostatic volume is important in determining the degree of hyperplastic enlargement, the resultant tendency toward urinary-tract outflow obstruction, and the preferred surgical treatment option. There are many medical imaging technologies to estimate prostate volume. Widely used technologies are Magnetic Resonance Imaging (MRI), Computed Tomography (CT), and Ultrasound (US). US technology differs from others with its portability, low- cost, and harmlessness, and it allows experts to scan the prostate in real time.
Trans Rectal Ultrasound (TRUS) and Abdominal Ultrasound (AUS) technologies are frequently used in prostate applications. Despite its better imaging quality with a higher Signal-to-Noise Ratio (SNR) and a larger view of the prostate with no other anatomic structures, TRUS technology is difficult to use regularly due to patient discomfort. The AUS technique is an easy-to-use alternative to imaging technology and is often used where TRUS is not practical.
Conventionally, a prostate-volume measurement is done manually on medical images by experts. Manual volume estimation results in high intra-expert and inter-expert difference due to factors caused by imaging quality, personal experience, and human error, which suggests that the guidance of experts by automatic systems would be beneficial. Automated prostate-volume-estimation systems are also essential to reduce the time spent measuring prostate volume.
The goal of this assignment is to develop a deep-learning-based algorithm for an accurate estimation of prostate volume using transabdominal ultrasound images.
