W-Net: Dense Semantic Segmentation of Subcutaneous Tissue in Ultrasound Images by Expanding U-Net to Incorporate Ultrasound RF Waveform Data
Authors:
Gautam Rajendrakumar Gare,
Jiayuan Li,
Rohan Joshi,
Mrunal Prashant Vaze,
Rishikesh Magar,
Michael Yousefpour,
Ricardo Luis Rodriguez,
John Micheal Galeotti
Abstract:
We present W-Net, a novel Convolution Neural Network (CNN) framework that employs raw ultrasound waveforms from each A-scan, typically referred to as ultrasound Radio Frequency (RF) data, in addition to the gray ultrasound image to semantically segment and label tissues. Unlike prior work, we seek to label every pixel in the image, without the use of a background class. To the best of our knowledg…
▽ More
We present W-Net, a novel Convolution Neural Network (CNN) framework that employs raw ultrasound waveforms from each A-scan, typically referred to as ultrasound Radio Frequency (RF) data, in addition to the gray ultrasound image to semantically segment and label tissues. Unlike prior work, we seek to label every pixel in the image, without the use of a background class. To the best of our knowledge, this is also the first deep-learning or CNN approach for segmentation that analyses ultrasound raw RF data along with the gray image. International patent(s) pending [PCT/US20/37519]. We chose subcutaneous tissue (SubQ) segmentation as our initial clinical goal since it has diverse intermixed tissues, is challenging to segment, and is an underrepresented research area. SubQ potential applications include plastic surgery, adipose stem-cell harvesting, lymphatic monitoring, and possibly detection/treatment of certain types of tumors. A custom dataset consisting of hand-labeled images by an expert clinician and trainees are used for the experimentation, currently labeled into the following categories: skin, fat, fat fascia/stroma, muscle and muscle fascia. We compared our results with U-Net and Attention U-Net. Our novel \emph{W-Net}'s RF-Waveform input and architecture increased mIoU accuracy (averaged across all tissue classes) by 4.5\% and 4.9\% compared to regular U-Net and Attention U-Net, respectively. We present analysis as to why the Muscle fascia and Fat fascia/stroma are the most difficult tissues to label. Muscle fascia in particular, the most difficult anatomic class to recognize for both humans and AI algorithms, saw mIoU improvements of 13\% and 16\% from our W-Net vs U-Net and Attention U-Net respectively.
△ Less
Submitted 2 September, 2020; v1 submitted 27 August, 2020;
originally announced August 2020.
