Deep Learning-Based Total Kidney Volume Segmentation in Autosomal Dominant Polycystic Kidney Disease Using Attention, Cosine Loss, and Sharpness Aware Minimization

Abstract

Early detection of the autosomal dominant polycystic kidney disease (ADPKD) is crucial as it is one of the most common causes of end-stage renal disease (ESRD) and kidney failure. The total kidney volume (TKV) can be used as a biomarker to quantify disease progression. The TKV calculation requires accurate delineation of kidney volumes, which is usually performed manually by an expert physician. However, this is time-consuming and automated segmentation is warranted. Furthermore, the scarcity of large annotated datasets hinders the development of deep learning solutions. In this work, we address this problem by implementing three attention mechanisms into the U-Net to improve TKV estimation. Additionally, we implement a cosine loss function that works well on image classification tasks with small datasets. Lastly, we apply a technique called sharpness aware minimization (SAM) that helps improve the generalizability of networks. Our results show significant improvements (p-value < 0.05) over the reference kidney segmentation U-Net. We show that the attention mechanisms and/or the cosine loss with SAM can achieve a dice score (DSC) of 0.918, a mean symmetric surface distance (MSSD) of 1.20 mm with the mean TKV difference of −1.72%, and R2 of 0.96 while using only 100 MRI datasets for training and testing. Furthermore, we tested four ensembles and obtained improvements over the best individual network, achieving a DSC and MSSD of 0.922 and 1.09 mm, respectively.

Deep Learning Automation of Kidney, Liver, and Spleen Segmentation for Organ Volume Measurements in Autosomal Dominant Polycystic Kidney Disease

“…This is especially important for MRI where minor improvements in the imaging technology are periodically implemented and need to be incorporated into the model. Unlike prior reports attempting to fully automate organ volume measurements into ADPKD with accuracy approaching manual contouring [22][23][24][25][26][27][28][29][30][31][32][33][34][35], this research demonstrates superior measurement reproducibility over manual contouring that can readily adapt to technological advances. Since the deep learning server is within the PACS firewall, technologists can rapidly transfer images to the server for running the inference.…”

“…This eliminates the need to manually draw every contour of the cystic kidneys, [ 22 ] thereby increasing the efficiency of accurate TKV measurement. Table 1 summarizes the existing literature for deep learning-based organ volume measurements in ADPKD using CT [ 23 , 24 , 25 , 26 , 27 ], ultrasound [ 28 ] and MRI [ 22 , 29 , 30 , 31 , 32 , 33 , 34 , 35 ]. MRI has the advantage over CT of not requiring ionizing radiation, which is particularly important, for these organ volume measurements are repeated many times over the patient’s lifetime, and MRI has higher resolution compared to ultrasound.…”

Effectiveness of Deep Learning Classifiers in Histopathological Diagnosis of Oral Squamous Cell Carcinoma by Pathologists

“…The best model utilized the SAM optimizer with VGG16 and a learning rate scheduler, as mentioned above. The SAM optimizer has been recently reported as a deep learning optimization method that performed well for publicly available datasets 10 , and classi ers using medical images 11,12 . Similar results were obtained using other deep learning classi ers research in this study.…”

Are We There Yet? The Value of Deep Learning in a Multicenter Setting for Response Prediction of Locally Advanced Rectal Cancer to Neoadjuvant Chemoradiotherapy

“…Furthermore, to mitigate artifacts related to image inhomogeneity, much larger datasets are required for network training [ 49 , 50 , 51 ]. Yet, the availability of image data seems to be one of the biggest hurdles for the implementation of artificial intelligence in the clinical setting [ 52 ], as supported by our study.…”