Search

• This paper proposes a novel architecture of a deep convolutional neural network consisting of a shared encoder and three parallel decoders specialized in kidney detection, renal parenchyma and cortex segmentation in contrast enhanced magnetic resonance imaging (CE-MRI). • The proposed network was validated in the leave-one-subject-out manner based on the cohort of 15 patients with diagnosed renal artery stenosis and 10 healthy volunteers. Healthy volunteers were scanned twice using the dynamic CE-MRI method. In between of the scanning sessions, renal performance of each subject was evaluated using iohexol clearance procedure to establish the ground-truth value of the glomerular filtration rate (GFR) and enable validation of the image-derived GFR measurements. • The obtained segmentation accuracy in terms of the Jaccard coefficient amounted to 94% for the whole kidney, and 76-92% for the renal cortex depending on the level of kidney tissue atrophy due to renal artery stenosis. • The average calculated volume of the whole kidney and renal cortex in the group of healthy subjects amounted to 160 and 102 cm3 respectively, as determined using the automatically found renal segments. The whole kidney volume decreases with the advancement of renal artery stenosis from 130 (mild stage) to 86 cm3 (severe). Similarly, the average renal cortex volume decreased from 106 to 69 cm3. • Repeatability of whole kidney and renal cortex volume measurements (healthy subjects) based on the proposed automatic segmentation procedure was in the range between 5.2% and 9.9% of the coefficient of variation. Repeatability of the image-derived single-kidney GFR measurements for the left and right kidney were correspondingly 15% and 21.0%, and stability improves in relation to calculations based on manual segmentations. Contrast-enhanced magnetic resonance imaging (CE-MRI) is one of the methods routinely used in clinics for the diagnosis of renal impairments. It allows assessment of kidney perfusion and also visualization of various lesions and tissue atrophy due to e.g. renal artery stenosis (RAS). An important indicator of the renal tissue state is the volume and shape of the kidney. Therefore it is highly desirable to equip radiological units in clinics with the software capable of automatic segmentation of the kidneys in CE-MRI images. This paper proposes a solution to this task using an original architecture of a deep neural network. The proposed design employs a three-branch convolutional neural network specialized in: 1) detection of renal parenchyma within an MR image patch, 2) segmentation of the whole kidney and 3) annotation of the renal cortex. We tested our architecture for normal kidneys in healthy subjects and for poorly perfused organs in RAS patients. The accuracy of renal parenchyma segmentation was equal to 0.94 in terms of the intersection over union (IoU) ratio. Accuracy of the cortex segmentation depends on the level of tissue health condition and ranges from 0.76 up to 0.92 of IoU. [ABSTRACT FROM AUTHOR]