Renal volume reconstruction using free-hand ultrasound scans

Kidney volume is a predictor of kidney function or the number of surviving nephrons; as such, it is a valuable biomarker in tracking the onset and progression of chronic kidney disease (CKD). Existing techniques to measure renal volumes [CT, MRI, three-dimensional (3D) ultrasound] are limited by high cost, long scan times, and low portability. To overcome these limitations, we propose a real-time, low-cost solution for estimating renal volumes using free-hand ultrasound scans. The method involves an Intel RealSense D415 camera mounted on a conventional one-dimensional ultrasound probe; RGB-D SLAM is used to localize the probe in free space. The acquired two-dimensional images are manually segmented by trained clinicians and combined with their corresponding poses to form a 3D volume. The method was tested on ex vivo sheep kidneys embedded in gelatin phantoms. Three different scanning protocols were tested: transverse linear scan (TL), transverse fan scan (TF), and longitudinal fan scan (LF). Ultrasound measured renal volumes were compared to those obtained using the water displacement (WD) method; TL = 66.2 ml, LF = 58.3 ml, TF = 53.3 ml, and WD = 66.6 ml. Further investigation includes automating kidney segmentation, accounting for patient motion, and in vivo validation.