The intersection of video capsule endoscopy and artificial intelligence: addressing unique challenges using machine learning.

Introduction: Technical burdens and time-intensive review processes limit the practical utility of video capsule endoscopy (VCE). Artificial intelligence (AI) is poised to address these limitations, but the intersection of AI and VCE reveals challenges that must first be overcome. We identified five challenges to address. Challenge #1: VCE data are stochastic and contains significant artifact. Challenge #2: VCE interpretation is cost-intensive. Challenge #3: VCE data are inherently imbalanced. Challenge #4: Existing VCE AIMLT are computationally cumbersome. Challenge #5: Clinicians are hesitant to accept AIMLT that cannot explain their process. Methods: An anatomic landmark detection model was used to test the application of convolutional neural networks (CNNs) to the task of classifying VCE data. We also created a tool that assists in expert annotation of VCE data. We then created more elaborate models using different approaches including a multi-frame approach, a CNN based on graph representation, and a few-shot approach based on meta-learning. Results: When used on full-length VCE footage, CNNs accurately identified anatomic landmarks (99.1%), with gradient weighted-class activation mapping showing the parts of each frame that the CNN used to make its decision. The graph CNN with weakly supervised learning (accuracy 89.9%, sensitivity of 91.1%), the few-shot model (accuracy 90.8%, precision 91.4%, sensitivity 90.9%), and the multi-frame model (accuracy 97.5%, precision 91.5%, sensitivity 94.8%) performed well. Discussion: Each of these five challenges is addressed, in part, by one of our AI-based models. Our goal of producing high performance using lightweight models that aim to improve clinician confidence was achieved.

Deep Learning Methods for Anatomical Landmark Detection in Video Capsule Endoscopy Images.

Video capsule endoscope (VCE) is an emerging technology that allows examination of the entire gastrointestinal (GI) tract with minimal invasion. While traditional endoscopy with biopsy procedures are the gold standard for diagnosis of most GI diseases, they are limited by how far the scope can be advanced in the tract and are also invasive. VCE allows gastroenterologists to investigate GI tract abnormalities in detail with visualization of all parts of the GI tract. It captures continuous real time images as it is propelled in the GI tract by gut motility. Even though VCE allows for thorough examination, reviewing and analyzing up to eight hours of images (compiled as videos) is tedious and not cost effective. In order to pave way for automation of VCE-based GI disease diagnosis, detecting the location of the capsule would allow for a more focused analysis as well as abnormality detection in each region of the GI tract. In this paper, we compared four deep Convolutional Neural Network models for feature extraction and detection of the anatomical part within the GI tract captured by VCE images. Our results showed that VGG-Net has superior performance with the highest average accuracy, precision, recall and, F1-score compared to other state of the art architectures: GoogLeNet, AlexNet and, ResNet.

Nephron number and its determinants: a 2020 update.

Studies of human nephron number have been conducted for well over a century and have uncovered a large variability in nephron number. However, the mechanisms influencing nephron endowment and loss, along with the etiology for the wide range among individuals are largely unknown. Advances in imaging technology have allowed investigators to revisit the principles of renal structure and physiology and their roles in the progression of kidney disease. Here, we will review the latest data on the influences impacting nephron number, innovations made over the last 6 years to understand and integrate renal structure and function, and new developments in the tools used to count nephrons in vivo.

Nephron loss detected by MRI following neonatal acute kidney injury in rabbits.

BACKGROUND: Acute kidney injury affects nearly 30% of preterm neonates in the intensive care unit. We aimed to determine whether nephrotoxin-induced AKI disrupted renal development assessed by imaging (CFE-MRI). METHODS: Neonatal New Zealand rabbits received indomethacin and gentamicin (AKI) or saline (control) for four days followed by cationic ferritin (CF) at six weeks. Ex vivo images were acquired using a gradient echo pulse sequence on 7 T MRI. Glomerular number (Nglom) and apparent glomerular volume (aVglom) were determined. CF toxicity was assessed at two and 28 days in healthy rabbits. RESULTS: Nglom was lower in the AKI group as compared to controls (74,034 vs 198,722, p < 0.01). aVglom was not different (AKI: 7.3 × 10-4 vs control: 6.2 × 10-4 mm3, p = 0.69). AKI kidneys had a band of glomeruli distributed radially in the cortex that were undetectable by MRI. Following CF injection, there was no difference in body or organ weights except for the liver, and transient changes in serum iron, platelets and white blood cell count. CONCLUSIONS: Brief nephrotoxin exposure during nephrogenesis results in fewer glomeruli and glomerular maldevelopment in a unique pattern detectable by MRI. Whole kidney evaluation by CFE-MRI may provide an important tool to understand the development of CKD following AKI.

Delayed Umbilical Cord Clamping is Not Associated with Acute Kidney Injury in Very Low Birth Weight Neonates.

OBJECTIVE: This study aimed to determine if delayed cord clamping (DCC) is associated with a reduction in neonatal acute kidney injury (AKI). STUDY DESIGN: A retrospective single-center cohort study of 278 very low birth weight (VLBW) neonates was performed to compare the incidence of AKI in the following groups: immediate cord clamping (ICC), DCC, and umbilical cord milking. AKI was diagnosed by the modified neonatal Kidney Diseases and Improving Global Outcomes (KDIGO) definition. RESULTS: The incidence of AKI in the first week was 20.1% with no difference between groups (p = 0.78). After adjustment for potential confounders, the odds of developing AKI, following DCC, compared with ICC was 0.93 (confidence interval [CI]: 0.46-1.86) with no reduction in the stage of AKI between groups. CONCLUSION: In this study, DCC was not associated with a reduced rate of AKI in VLBW neonates. However, the data suggest that DCC is also not harmful to the kidneys, further supporting the safety of DCC in VLBW neonates.

Is acute kidney injury a harbinger for chronic kidney disease?

PURPOSE OF REVIEW: Despite abundant evidence in adults, the relationship between acute kidney injury (AKI) and chronic kidney disease (CKD) remains unanswered in pediatrics. Obstacles to overcome include the challenges defining these entities and the lack of long-term follow-up studies. This review focuses on pediatric populations at high-risk for AKI, the evidence of the long-term effect of AKI on renal health, and biomarkers to detect renal disease. RECENT FINDINGS: AKI in critically ill children and neonates is common and independently associated with adverse outcomes. Patients with diabetes and sickle cell disease along with neonates with necrotizing enterocolitis have been identified as high-risk for AKI. Preterm birth and neonates with AKI have signs of renal dysfunction early in childhood. Urinary biomarkers may identify AKI and CKD earlier than traditional biomarkers, but more work is necessary to determine their clinical utility. Promising technological advances including the ability to determine nephron number noninvasively will expand our ability to characterize the AKI to CKD transition. SUMMARY: AKI is common and associated with poor outcomes. It is probable that AKI is a harbinger to CKD in pediatric populations. However, we currently lack the tools to definitely answer this question and more research is needed.

Lipid-dependent regulation of exocytosis in S. cerevisiae by OSBP homolog (Osh) 4.

Polarized exocytosis is an essential process in many organisms and cell types for correct cell division or functional specialization. Previous studies established that homologs of the oxysterol-binding protein (OSBP) in S. cerevisiae, which comprise the Osh protein family, are necessary for efficient polarized exocytosis by supporting a late post-Golgi step. We define this step as the docking of a specific sub-population of exocytic vesicles with the plasma membrane. In the absence of other Osh proteins, yeast Osh4p can support this process in a manner dependent upon two lipid ligands, PI4P and sterol. Osh6p, which binds PI4P and phosphatidylserine, is also sufficient to support polarized exocytosis, again in a lipid-dependent manner. These data suggest that Osh-mediated exocytosis depends upon lipid binding and exchange without a strict requirement for sterol. We propose a two-step mechanism for Osh protein-mediated regulation of polarized exocytosis by using Osh4p as a model. We describe a specific in vivo role for lipid binding by an OSBP-related protein (ORP) in the process of polarized exocytosis, guiding our understanding of where and how OSBP and ORPs may function in more complex organisms.