Deployed Deep Learning Kidney Segmentation for Polycystic Kidney Disease MRI

Deep Learning-Based Automated Imaging Classification of ADPKD

Introduction

The Mayo imaging classification model (MICM) requires a prestep qualitative assessment to determine whether a patient is in class 1 (typical) or class 2 (atypical), where patients assigned to class 2 are excluded from the MICM application.

Methods

We developed a deep learning-based method to automatically classify class 1 and 2 from magnetic resonance (MR) images and provide classification confidence utilizing abdominal T 2 -weighted MR images from 486 subjects, where transfer learning was applied. In addition, the explainable artificial intelligence (XAI) method was illustrated to enhance the explainability of the automated classification results. For performance evaluations, confusion matrices were generated, and receiver operating characteristic curves were drawn to measure the area under the curve.

Results

The proposed method showed excellent performance for the classification of class 1 (97.7%) and 2 (100%), where the combined test accuracy was 98.01%. The precision and recall for predicting class 1 were 1.00 and 0.98, respectively, with F 1 -score of 0.99; whereas those for predicting class 2 were 0.87 and 1.00, respectively, with F 1 -score of 0.93. The weighted averages of precision and recall were 0.98 and 0.98, respectively, showing the classification confidence scores whereas the XAI method well-highlighted contributing regions for the classification.

Conclusion

The proposed automated method can classify class 1 and 2 cases as accurately as the level of a human expert. This method may be a useful tool to facilitate clinical trials investigating different types of kidney morphology and for clinical management of patients with autosomal dominant polycystic kidney disease (ADPKD).

Kidney medicine meets computer vision: a bibliometric analysis

BACKGROUND AND OBJECTIVE

Rapid advances in computer vision (CV) have the potential to facilitate the examination, diagnosis, and treatment of diseases of the kidney. The bibliometric study aims to explore the research landscape and evolving research focus of the application of CV in kidney medicine research.

METHODS

The Web of Science Core Collection was utilized to identify publications related to the research or applications of CV technology in the field of kidney medicine from January 1, 1900, to December 31, 2022. We analyzed emerging research trends, highly influential publications and journals, prolific researchers, countries/regions, research institutions, co-authorship networks, and co-occurrence networks. Bibliographic information was analyzed and visualized using Python, Matplotlib, Seaborn, HistCite, and Vosviewer.

RESULTS

There was an increasing trend in the number of publications on CV-based kidney medicine research. These publications mainly focused on medical image processing, surgical procedures, medical image analysis/diagnosis, as well as the application and innovation of CV technology in medical imaging. The United States is currently the leading country in terms of the quantities of published articles and international collaborations, followed by China. Deep learning-based segmentation and machine learning-based texture analysis are the most commonly used techniques in this field. Regarding research hotspot trends, CV algorithms are shifting toward artificial intelligence, and research objects are expanding to encompass a wider range of kidney-related objects, with data dimensions used in research transitioning from 2D to 3D while simultaneously incorporating more diverse data modalities.

CONCLUSION

The present study provides a scientometric overview of the current progress in the research and application of CV technology in kidney medicine research. Through the use of bibliometric analysis and network visualization, we elucidate emerging trends, key sources, leading institutions, and popular topics. Our findings and analysis are expected to provide valuable insights for future research on the use of CV in kidney medicine research.

A Primer for Utilizing Deep Learning and Abdominal MRI Imaging Features to Monitor Autosomal Dominant Polycystic Kidney Disease Progression

Abdominal imaging of autosomal dominant polycystic kidney disease (ADPKD) has historically focused on detecting complications such as cyst rupture, cyst infection, obstructing renal calculi, and pyelonephritis; discriminating complex cysts from renal cell carcinoma; and identifying sources of abdominal pain. Many imaging features of ADPKD are incompletely evaluated or not deemed to be clinically significant, and because of this, treatment options are limited. However, total kidney volume (TKV) measurement has become important for assessing the risk of disease progression (i.e., Mayo Imaging Classification) and predicting tolvaptan treatment's efficacy. Deep learning for segmenting the kidneys has improved these measurements' speed, accuracy, and reproducibility. Deep learning models can also segment other organs and tissues, extracting additional biomarkers to characterize the extent to which extrarenal manifestations complicate ADPKD. In this concept paper, we demonstrate how deep learning may be applied to measure the TKV and how it can be extended to measure additional features of this disease.

Dual-task kidney MR segmentation with transformers in autosomal-dominant polycystic kidney disease

Autosomal-dominant polycystic kidney disease is a prevalent genetic disorder characterized by the development of renal cysts, leading to kidney enlargement and renal failure. Accurate measurement of total kidney volume through polycystic kidney segmentation is crucial to assess disease severity, predict progression and evaluate treatment effects. Traditional manual segmentation suffers from intra- and inter-expert variability, prompting the exploration of automated approaches. In recent years, convolutional neural networks have been employed for polycystic kidney segmentation from magnetic resonance images. However, the use of Transformer-based models, which have shown remarkable performance in a wide range of computer vision and medical image analysis tasks, remains unexplored in this area. With their self-attention mechanism, Transformers excel in capturing global context information, which is crucial for accurate organ delineations. In this paper, we evaluate and compare various convolutional-based, Transformers-based, and hybrid convolutional/Transformers-based networks for polycystic kidney segmentation. Additionally, we propose a dual-task learning scheme, where a common feature extractor is followed by per-kidney decoders, towards better generalizability and efficiency. We extensively evaluate various architectures and learning schemes on a heterogeneous magnetic resonance imaging dataset collected from 112 patients with polycystic kidney disease. Our results highlight the effectiveness of Transformer-based models for polycystic kidney segmentation and the relevancy of exploiting dual-task learning to improve segmentation accuracy and mitigate data scarcity issues. A promising ability in accurately delineating polycystic kidneys is especially shown in the presence of heterogeneous cyst distributions and adjacent cyst-containing organs. This work contribute to the advancement of reliable delineation methods in nephrology, paving the way for a broad spectrum of clinical applications.

Test Retest Reproducibility of Organ Volume Measurements in ADPKD Using 3D Multimodality Deep Learning

RATIONALE AND OBJECTIVES

Following autosomal dominant polycystic kidney disease (ADPKD) progression by measuring organ volumes requires low measurement variability. The objective of this study is to reduce organ volume measurement variability on MRI of ADPKD patients by utilizing all pulse sequences to obtain multiple measurements which allows outlier analysis to find errors and averaging to reduce variability.

MATERIALS AND METHODS

In order to make measurements on multiple pulse sequences practical, a 3D multi-modality multi-class segmentation model based on nnU-net was trained/validated using T1, T2, SSFP, DWI and CT from 413 subjects. Reproducibility was assessed with test-re-test methodology on ADPKD subjects (n = 19) scanned twice within a 3-week interval correcting outliers and averaging the measurements across all sequences. Absolute percent differences in organ volumes were compared to paired students t-test.

RESULTS

Dice similarlity coefficient > 97%, Jaccard Index > 0.94, mean surface distance < 1 mm and mean Hausdorff Distance < 2 cm for all three organs and all five sequences were found on internal (n = 25), external (n = 37) and test-re-test reproducibility assessment (38 scans in 19 subjects). When averaging volumes measured from five MRI sequences, the model automatically segmented kidneys with test-re-test reproducibility (percent absolute difference between exam 1 and exam 2) of 1.3% which was better than all five expert observers. It reliably stratified ADPKD into Mayo Imaging Classification (area under the curve=100%) compared to radiologist.

CONCLUSION

3D deep learning measures organ volumes on five MRI sequences leveraging the power of outlier analysis and averaging to achieve 1.3% total kidney test-re-test reproducibility.

AI models for automated segmentation of engineered polycystic kidney tubules

Autosomal dominant polycystic kidney disease (ADPKD) is a monogenic, rare disease, characterized by the formation of multiple cysts that grow out of the renal tubules. Despite intensive attempts to develop new drugs or repurpose existing ones, there is currently no definitive cure for ADPKD. This is primarily due to the complex and variable pathogenesis of the disease and the lack of models that can faithfully reproduce the human phenotype. Therefore, the development of models that allow automated detection of cysts' growth directly on human kidney tissue is a crucial step in the search for efficient therapeutic solutions. Artificial Intelligence methods, and deep learning algorithms in particular, can provide powerful and effective solutions to such tasks, and indeed various architectures have been proposed in the literature in recent years. Here, we comparatively review state-of-the-art deep learning segmentation models, using as a testbed a set of sequential RGB immunofluorescence images from 4 in vitro experiments with 32 engineered polycystic kidney tubules. To gain a deeper understanding of the detection process, we implemented both pixel-wise and cyst-wise performance metrics to evaluate the algorithms. Overall, two models stand out as the best performing, namely UNet++ and UACANet: the latter uses a self-attention mechanism introducing some explainability aspects that can be further exploited in future developments, thus making it the most promising algorithm to build upon towards a more refined cyst-detection platform. UACANet model achieves a cyst-wise Intersection over Union of 0.83, 0.91 for Recall, and 0.92 for Precision when applied to detect large-size cysts. On all-size cysts, UACANet averages at 0.624 pixel-wise Intersection over Union. The code to reproduce all results is freely available in a public GitHub repository.

Intra- and Inter-expert Validation of an Automatic Segmentation Method for Fluid Regions Associated with Central Serous Chorioretinopathy in OCT Images

Central Serous Chorioretinopathy (CSC) is a retinal disorder caused by the accumulation of fluid, resulting in vision distortion. The diagnosis of this disease is typically performed through Optical Coherence Tomography (OCT) imaging, which displays any fluid buildup between the retinal layers. Currently, these fluid regions are manually detected by visual inspection a time-consuming and subjective process that can be prone to errors. A series of six deep learning-based automatic segmentation architectural configurations of different levels of complexity were trained and compared in order to determine the best model intended for the automatic segmentation of CSC-related lesions in OCT images. The best performing models were then evaluated in an external validation study. Furthermore, an intra- and inter-expert analysis was conducted in order to compare the manual segmentation performed by expert ophthalmologists with the automatic segmentation provided by the models. Test results of the best performing configuration achieved a mean Dice of 0.868 ± 0.056 in the internal dataset. In the external validation set, these models achieved a level of agreement with human experts of up to 0.960 in terms of Kappa coefficient, contrasting with a value of 0.951 for agreement between human experts. Overall, the models reached a better agreement with either of the human experts than these experts with each other, suggesting that automatic segmentation models for the detection of CSC-related lesions in OCT imaging can be useful tools for assessing this disease, reducing the workload of manual inspection and leading to a more robust and objective diagnosis method.

Multiparametric MRI-Based Machine Learning Models for the Characterization of Cystic Renal Masses Compared to the Bosniak Classification, Version 2019: A Multicenter Study

RATIONALE AND OBJECTIVE

Accurate differentiation between benign and malignant cystic renal masses (CRMs) is challenging in clinical practice. This study aimed to develop MRI-based machine learning models for differentiating between benign and malignant CRMs and compare the best-performing model with the Bosniak classification, version 2019 (BC, version 2019).

METHODS

Between 2009 and 2021, consecutive surgery-proven CRM patients with renal MRI were enrolled in this multicenter study. Models were constructed to differentiate between benign and malignant CRMs using logistic regression (LR), random forest (RF), and support vector machine (SVM) algorithms, respectively. Meanwhile, two radiologists classified CRMs into I-IV categories according to the BC, version 2019 in consensus in the test set. A subgroup analysis was conducted to investigate the performance of the best-performing model in complicated CRMs (II-IV lesions in the test set). The performances of models and BC, version 2019 were evaluated using the area under the receiver operating characteristic curve (AUC). Performance was statistically compared between the best-performing model and the BC, version 2019.

RESULTS

278 and 48 patients were assigned to the training and test sets, respectively. In the test set, the AUC and accuracy of the LR model, the RF model, the SVM model, and the BC, version 2019 were 0.884 and 75.0%, 0.907 and 83.3%, 0.814 and 72.9%, and 0.893 and 81.2%, respectively. Neither the AUC nor the accuracy of the RF model that performed best were significantly different from the BC, version 2019 (P = 0.780, P = 0.065). The RF model achieved an AUC and accuracy of 0.880 and 81.0% in complicated CRMs.

CONCLUSIONS

The MRI-based RF model can accurately differentiate between benign and malignant CRMs with comparable performance to the BC, version 2019, and has good performance in complicated CRMs, which may facilitate treatment decision-making and is less affected by interobserver disagreements.

Total kidney volume in autosomal dominant polycystic kidney disease: intraobserver and interobserver agreement of two methods with MRI

Background/aim

Total kidney volume (TKV) is a parameter used in both treatment decision and follow-up in autosomal dominant polycystic kidney disease (ADPKD) patients. The objective of this study was to evaluate intra- and interobserver agreement of the ellipsoid formula (EF) and manual boundary tracing method (MBTM) used in TKV measurement of ADPKD patients across different levels of experience radiologists. Additionally, the study aimed to evaluate the correlation between the EF and MBTM, which is considered the gold standard for TKV.

Materials and methods

A retrospective evaluation was conducted on magnetic resonance imaging (MRI) data from 55 ADPKD patients who underwent abdominal MRI between January 2017 and November 2021 to evaluate TKV. TKV measurements were performed by three independent observers (observer 1, an abdominal imaging radiologist with 5 years of experience; observer 2, a fourth-year radiology resident; observer 3, a second-year radiology resident).To assess intraobserver variability, all observers repeated the measurements at two-week intervals. The ICC was used to assess both intraobserver and interobserver variability. A comparison of the two methods was performed by linear regression for all three observers.

Results

The ICC (95% CI) indicated excellent agreement between the observers for both methods (among all observers, p < 0.001). Furthermore, excellent intraobserver agreement was found between all observer measurements either EF or MBTM based on ICC (95% CI) (p < 0.001). The results of the linear regression analysis demonstrated high correlations between the two methods in all three observers (r = 0.992, p < 0.001 for the first observer; r = 0.975, p < 0.001 for the second observer; r = 0.989, p < 0.001 for the third observer).

Conclusion

Both the EF and MBTM methods used for the measurement of TKV provided excellent intra- and interobserver reproducibility. The EF is as accurate and precise as the MBTM. It may therefore be preferred in radiology departments with heavy workload, as it is a reliable method for rapid and easy assessment, independent of experience.

Deep learning-based segmentation of multisite disease in ovarian cancer

PURPOSE

To determine if pelvic/ovarian and omental lesions of ovarian cancer can be reliably segmented on computed tomography (CT) using fully automated deep learning-based methods.

METHODS

A deep learning model for the two most common disease sites of high-grade serous ovarian cancer lesions (pelvis/ovaries and omentum) was developed and compared against the well-established "no-new-Net" framework and unrevised trainee radiologist segmentations. A total of 451 CT scans collected from four different institutions were used for training (n = 276), evaluation (n = 104) and testing (n = 71) of the methods. The performance was evaluated using the Dice similarity coefficient (DSC) and compared using a Wilcoxon test.

RESULTS

Our model outperformed no-new-Net for the pelvic/ovarian lesions in cross-validation, on the evaluation and test set by a significant margin (p values being 4 × 10-7, 3 × 10-4, 4 × 10-2, respectively), and for the omental lesions on the evaluation set (p = 1 × 10-3). Our model did not perform significantly differently in segmenting pelvic/ovarian lesions (p = 0.371) compared to a trainee radiologist. On an independent test set, the model achieved a DSC performance of 71 ± 20 (mean ± standard deviation) for pelvic/ovarian and 61 ± 24 for omental lesions.

CONCLUSION

Automated ovarian cancer segmentation on CT scans using deep neural networks is feasible and achieves performance close to a trainee-level radiologist for pelvic/ovarian lesions.

RELEVANCE STATEMENT

Automated segmentation of ovarian cancer may be used by clinicians for CT-based volumetric assessments and researchers for building complex analysis pipelines.

KEY POINTS

• The first automated approach for pelvic/ovarian and omental ovarian cancer lesion segmentation on CT images has been presented. • Automated segmentation of ovarian cancer lesions can be comparable with manual segmentation of trainee radiologists. • Careful hyperparameter tuning can provide models significantly outperforming strong state-of-the-art baselines.

Developing and deploying deep learning models in brain magnetic resonance imaging: A review

Magnetic resonance imaging (MRI) of the brain has benefited from deep learning (DL) to alleviate the burden on radiologists and MR technologists, and improve throughput. The easy accessibility of DL tools has resulted in a rapid increase of DL models and subsequent peer-reviewed publications. However, the rate of deployment in clinical settings is low. Therefore, this review attempts to bring together the ideas from data collection to deployment in the clinic, building on the guidelines and principles that accreditation agencies have espoused. We introduce the need for and the role of DL to deliver accessible MRI. This is followed by a brief review of DL examples in the context of neuropathologies. Based on these studies and others, we collate the prerequisites to develop and deploy DL models for brain MRI. We then delve into the guiding principles to develop good machine learning practices in the context of neuroimaging, with a focus on explainability. A checklist based on the United States Food and Drug Administration's good machine learning practices is provided as a summary of these guidelines. Finally, we review the current challenges and future opportunities in DL for brain MRI.

Renal MRI: From Nephron to NMR Signal

Renal diseases pose a significant socio-economic burden on healthcare systems. The development of better diagnostics and prognostics is well-recognized as a key strategy to resolve these challenges. Central to these developments are MRI biomarkers, due to their potential for monitoring of early pathophysiological changes, renal disease progression or treatment effects. The surge in renal MRI involves major cross-domain initiatives, large clinical studies, and educational programs. In parallel with these translational efforts, the need for greater (patho)physiological specificity remains, to enable engagement with clinical nephrologists and increase the associated health impact. The ISMRM 2022 Member Initiated Symposium (MIS) on renal MRI spotlighted this issue with the goal of inspiring more solutions from the ISMRM community. This work is a summary of the MIS presentations devoted to: 1) educating imaging scientists and clinicians on renal (patho)physiology and demands from clinical nephrologists, 2) elucidating the connection of MRI parameters with renal physiology, 3) presenting the current state of leading MR surrogates in assessing renal structure and functions as well as their next generation of innovation, and 4) describing the potential of these imaging markers for providing clinically meaningful renal characterization to guide or supplement clinical decision making. We hope to continue momentum of recent years and introduce new entrants to the development process, connecting (patho)physiology with (bio)physics, and conceiving new clinical applications. We envision this process to benefit from cross-disciplinary collaboration and analogous efforts in other body organs, but also to maximally leverage the unique opportunities of renal physiology. LEVEL OF EVIDENCE: 1 TECHNICAL EFFICACY STAGE: 2.

Convolutional neural network-based kidney volume estimation from low-dose unenhanced computed tomography scans

PURPOSE

Kidney volume is important in the management of renal diseases. Unfortunately, the currently available, semi-automated kidney volume determination is time-consuming and prone to errors. Recent advances in its automation are promising but mostly require contrast-enhanced computed tomography (CT) scans. This study aimed at establishing an automated estimation of kidney volume in non-contrast, low-dose CT scans of patients with suspected urolithiasis.

METHODS

The kidney segmentation process was automated with 2D Convolutional Neural Network (CNN) models trained on manually segmented 2D transverse images extracted from low-dose, unenhanced CT scans of 210 patients. The models' segmentation accuracy was assessed using Dice Similarity Coefficient (DSC), for the overlap with manually-generated masks on a set of images not used in the training. Next, the models were applied to 22 previously unseen cases to segment kidney regions. The volume of each kidney was calculated from the product of voxel number and their volume in each segmented mask. Kidney volume results were then validated against results semi-automatically obtained by radiologists.

RESULTS

The CNN-enabled kidney volume estimation took a mean of 32 s for both kidneys in a CT scan with an average of 1026 slices. The DSC was 0.91 and 0.86 and for left and right kidneys, respectively. Inter-rater variability had consistencies of ICC = 0.89 (right), 0.92 (left), and absolute agreements of ICC = 0.89 (right), 0.93 (left) between the CNN-enabled and semi-automated volume estimations.

CONCLUSION

In our work, we demonstrated that CNN-enabled kidney volume estimation is feasible and highly reproducible in low-dose, non-enhanced CT scans. Automatic segmentation can thereby quantitatively enhance radiological reports.

Effect of Averaging Measurements From Multiple MRI Pulse Sequences on Kidney Volume Reproducibility in Autosomal Dominant Polycystic Kidney Disease

BACKGROUND

Total kidney volume (TKV) is an important biomarker for assessing kidney function, especially for autosomal dominant polycystic kidney disease (ADPKD). However, TKV measurements from a single MRI pulse sequence have limited reproducibility, ± ~5%, similar to ADPKD annual kidney growth rates.

PURPOSE

To improve TKV measurement reproducibility on MRI by extending artificial intelligence algorithms to automatically segment kidneys on T1-weighted, T2-weighted, and steady state free precession (SSFP) sequences in axial and coronal planes and averaging measurements.

STUDY TYPE

Retrospective training, prospective testing.

SUBJECTS

Three hundred ninety-seven patients (356 with ADPKD, 41 without), 75% for training and 25% for validation, 40 ADPKD patients for testing and 17 ADPKD patients for assessing reproducibility.

FIELD STRENGTH/SEQUENCE

T2-weighted single-shot fast spin echo (T2), SSFP, and T1-weighted 3D spoiled gradient echo (T1) at 1.5 and 3T.

ASSESSMENT

2D U-net segmentation algorithm was trained on images from all sequences. Five observers independently measured each kidney volume manually on axial T2 and using model-assisted segmentations on all sequences and image plane orientations for two MRI exams in two sessions separated by 1-3 weeks to assess reproducibility. Manual and model-assisted segmentation times were recorded.

STATISTICAL TESTS

Bland-Altman, Schapiro-Wilk (normality assessment), Pearson's chi-squared (categorical variables); Dice similarity coefficient, interclass correlation coefficient, and concordance correlation coefficient for analyzing TKV reproducibility. P-value < 0.05 was considered statistically significant.

RESULTS

In 17 ADPKD subjects, model-assisted segmentations of axial T2 images were significantly faster than manual segmentations (2:49 minute vs. 11:34 minute), with no significant absolute percent difference in TKV (5.9% vs. 5.3%, P = 0.88) between scans 1 and 2. Absolute percent differences between the two scans for model-assisted segmentations on other sequences were 5.5% (axial T1), 4.5% (axial SSFP), 4.1% (coronal SSFP), and 3.2% (coronal T2). Averaging measurements from all five model-assisted segmentations significantly reduced absolute percent difference to 2.5%, further improving to 2.1% after excluding an outlier.

DATA CONCLUSION

Measuring TKV on multiple MRI pulse sequences in coronal and axial planes is practical with deep learning model-assisted segmentations and can improve TKV measurement reproducibility more than 2-fold in ADPKD.

EVIDENCE LEVEL

2 TECHNICAL EFFICACY: Stage 1.

"Hi, how can i help you?": embracing artificial intelligence in kidney research

In recent years, biology and precision medicine have benefited from major advancements in generating large-scale molecular and biomedical datasets and in analyzing those data using advanced machine learning algorithms. Machine learning applications in kidney physiology and pathophysiology include segmenting kidney structures from imaging data and predicting conditions like acute kidney injury or chronic kidney disease using electronic health records. Despite the potential of machine learning to revolutionize nephrology by providing innovative diagnostic and therapeutic tools, its adoption in kidney research has been slower than in other organ systems. Several factors contribute to this underutilization. The complexity of the kidney as an organ, with intricate physiology and specialized cell populations, makes it challenging to extrapolate bulk omics data to specific processes. In addition, kidney diseases often present with overlapping manifestations and morphological changes, making diagnosis and treatment complex. Moreover, kidney diseases receive less funding compared with other pathologies, leading to lower awareness and limited public-private partnerships. To promote the use of machine learning in kidney research, this review provides an introduction to machine learning and reviews its notable applications in renal research, such as morphological analysis, omics data examination, and disease diagnosis and prognosis. Challenges and limitations associated with data-driven predictive techniques are also discussed. The goal of this review is to raise awareness and encourage the kidney research community to embrace machine learning as a powerful tool that can drive advancements in understanding kidney diseases and improving patient care.

Abdominal Imaging in ADPKD: Beyond Total Kidney Volume

In the context of autosomal dominant polycystic kidney disease (ADPKD), measurement of the total kidney volume (TKV) is crucial. It acts as a marker for tracking disease progression, and evaluating the effectiveness of treatment strategies. The TKV has also been recognized as an enrichment biomarker and a possible surrogate endpoint in clinical trials. Several imaging modalities and methods are available to calculate the TKV, and the choice depends on the purpose of use. Technological advancements have made it possible to accurately assess the cyst burden, which can be crucial to assessing the disease state and helping to identify rapid progressors. Moreover, the development of automated algorithms has increased the efficiency of total kidney and cyst volume measurements. Beyond these measurements, the quantification and characterization of non-cystic kidney tissue shows potential for stratifying ADPKD patients early on, monitoring disease progression, and possibly predicting renal function loss. A broad spectrum of radiological imaging techniques are available to characterize the kidney tissue, showing promise when it comes to non-invasively picking up the early signs of ADPKD progression. Radiomics have been used to extract textural features from ADPKD images, providing valuable information about the heterogeneity of the cystic and non-cystic components. This review provides an overview of ADPKD imaging biomarkers, focusing on the quantification methods, potential, and necessary steps toward a successful translation to clinical practice.

Automated MRI Field of View Prescription from Region of Interest Prediction by Intra-Stack Attention Neural Network

Manual prescription of the field of view (FOV) by MRI technologists is variable and prolongs the scanning process. Often, the FOV is too large or crops critical anatomy. We propose a deep learning framework, trained by radiologists' supervision, for automating FOV prescription. An intra-stack shared feature extraction network and an attention network are used to process a stack of 2D image inputs to generate scalars defining the location of a rectangular region of interest (ROI). The attention mechanism is used to make the model focus on a small number of informative slices in a stack. Then, the smallest FOV that makes the neural network predicted ROI free of aliasing is calculated by an algebraic operation derived from MR sampling theory. The framework's performance is examined quantitatively with intersection over union (IoU) and pixel error on position and qualitatively with a reader study. The proposed model achieves an average IoU of 0.867 and an average ROI position error of 9.06 out of 512 pixels on 80 test cases, significantly better than two baseline models and not significantly different from a radiologist. Finally, the FOV given by the proposed framework achieves an acceptance rate of 92% from an experienced radiologist.

Deep Learning for Image Analysis in Kidney Care

Analysis of medical images, such as radiological or tissue specimens, is an indispensable part of medical diagnostics. Conventionally done manually, the process may sometimes be time-consuming and prone to interobserver variability. Image classification and segmentation by deep learning strategies, predominantly convolutional neural networks, may provide a significant advance in the diagnostic process. In renal medicine, most evidence has been generated around the radiological assessment of renal abnormalities and histological analysis of renal biopsy specimens' segmentation. In this article, the basic principles of image analysis by convolutional neural networks, brief descriptions of convolutional neural networks, and their system architecture for image analysis are discussed, in combination with examples regarding their use in image analysis in nephrology.

Automated Kidney and Liver Segmentation in MR Images in Patients with Autosomal Dominant Polycystic Kidney Disease: A Multicenter Study

Background

Imaging-based total kidney volume (TKV) and total liver volume (TLV) are major prognostic factors in autosomal dominant polycystic kidney disease (ADPKD) and end points for clinical trials. However, volumetry is time consuming and reader dependent in clinical practice. Our aim was to develop a fully automated method for joint kidney and liver segmentation in magnetic resonance imaging (MRI) and to evaluate its performance in a multisequence, multicenter setting.

Methods

The convolutional neural network was trained on a large multicenter dataset consisting of 992 MRI scans of 327 patients. Manual segmentation delivered ground-truth labels. The model's performance was evaluated in a separate test dataset of 93 patients (350 MRI scans) as well as a heterogeneous external dataset of 831 MRI scans from 323 patients.

Results

The segmentation model yielded excellent performance, achieving a median per study Dice coefficient of 0.92-0.97 for the kidneys and 0.96 for the liver. Automatically computed TKV correlated highly with manual measurements (intraclass correlation coefficient [ICC]: 0.996-0.999) with low bias and high precision (-0.2%±4% for axial images and 0.5%±4% for coronal images). TLV estimation showed an ICC of 0.999 and bias/precision of -0.5%±3%. For the external dataset, the automated TKV demonstrated bias and precision of -1%±7%.

Conclusions

Our deep learning model enabled accurate segmentation of kidneys and liver and objective assessment of TKV and TLV. Importantly, this approach was validated with axial and coronal MRI scans from 40 different scanners, making implementation in clinical routine care feasible.Clinical Trial registry name and registration number: The German ADPKD Tolvaptan Treatment Registry (AD[H]PKD), NCT02497521.

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

Organ volume measurements are a key metric for managing ADPKD (the most common inherited renal disease). However, measuring organ volumes is tedious and involves manually contouring organ outlines on multiple cross-sectional MRI or CT images. The automation of kidney contouring using deep learning has been proposed, as it has small errors compared to manual contouring. Here, a deployed open-source deep learning ADPKD kidney segmentation pipeline is extended to also measure liver and spleen volumes, which are also important. This 2D U-net deep learning approach was developed with radiologist labeled T2-weighted images from 215 ADPKD subjects (70% training = 151, 30% validation = 64). Additional ADPKD subjects were utilized for prospective (n = 30) and external (n = 30) validations for a total of 275 subjects. Image cropping previously optimized for kidneys was included in training but removed for the validation and inference to accommodate the liver which is closer to the image border. An effective algorithm was developed to adjudicate overlap voxels that are labeled as more than one organ. Left kidney, right kidney, liver and spleen labels had average errors of 3%, 7%, 3%, and 1%, respectively, on external validation and 5%, 6%, 5%, and 1% on prospective validation. Dice scores also showed that the deep learning model was close to the radiologist contouring, measuring 0.98, 0.96, 0.97 and 0.96 on external validation and 0.96, 0.96, 0.96 and 0.95 on prospective validation for left kidney, right kidney, liver and spleen, respectively. The time required for manual correction of deep learning segmentation errors was only 19:17 min compared to 33:04 min for manual segmentations, a 42% time saving (p = 0.004). Standard deviation of model assisted segmentations was reduced to 7, 5, 11, 5 mL for right kidney, left kidney, liver and spleen respectively from 14, 10, 55 and 14 mL for manual segmentations. Thus, deep learning reduces the radiologist time required to perform multiorgan segmentations in ADPKD and reduces measurement variability.