Renal cortex, medulla and pelvicaliceal system segmentation on arterial phase CT images with random patch-based networks

Tang, Yucheng; Gao, Riqiang; Lee, Ho Hin; Xu, Zhoubing; Savoie, Brent; Bao, Shunxing; Huo, Yuankai; Fogo, Agnes B.; Harris, Raymond C.; Caestecker, Mark P. de; Spraggins, Jeffrey M.; Landman, Bennett A.

doi:10.1117/12.2581101

Cited by 6 publications

(4 citation statements)

References 21 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Previous studies have already successfully segmented renal images [8][9][10][11][12][13] but they were restricted to the whole kidney volume or the volume occupied by renal cysts. Our approach extends the whole kidney segmentation, which is the first step of our proposed RCM U-Net, to the segmentation of both cortex and medulla separately.…”

Section: Plos Onementioning

confidence: 99%

See 1 more Smart Citation

Validation of automatically measured T1 map cortico-medullary difference (ΔT1) for eGFR and fibrosis assessment in allograft kidneys

et al. 2023

View full text Add to dashboard Cite

MRI T1-mapping is an important non-invasive tool for renal diagnosis. Previous work shows that ΔT1 (cortex-medullary difference in T1) has significant correlation with interstitial fibrosis in chronic kidney disease (CKD) allograft patients. However, measuring cortico-medullary values by manually drawing ROIs over cortex and medulla (a gold standard method) is challenging, time-consuming, subjective and requires human training. Moreover, such subjective ROI placement may also affect the work reproducibility. This work proposes a deep learning-based 2D U-Net (RCM U-Net) to auto-segment the renal cortex and medulla of CKD allograft kidney T1 maps. Furthermore, this study presents a correlation of automatically measured ΔT1 values with eGFR and percentage fibrosis in allograft kidneys. Also, the RCM U-Net correlation results are compared with the manual ROI correlation analysis. The RCM U-Net has been trained and validated on T1 maps from 40 patients (n = 2400 augmented images) and tested on 10 patients (n = 600 augmented images). The RCM U-Net segmentation results are compared with the standard VGG16, VGG19, ResNet34 and ResNet50 networks with U-Net as backbone. For clinical validation of the RCM U-Net segmentation, another set of 114 allograft kidneys patient’s cortex and medulla were automatically segmented to measure the ΔT1 values and correlated with eGFR and fibrosis. Overall, the RCM U-Net showed 50% less Mean Absolute Error (MAE), 16% better Dice Coefficient (DC) score and 12% improved results in terms of Sensitivity (SE) over conventional CNNs (i.e. VGG16, VGG19, ResNet34 and ResNet50) while the Specificity (SP) and Accuracy (ACC) did not show significant improvement (i.e. 0.5% improvement) for both cortex and medulla segmentation. For eGFR and fibrosis assessment, the proposed RCM U-Net correlation coefficient (r) and R-square (R2) was better correlated (r = -0.2, R2 = 0.041 with p = 0.039) to eGFR than manual ROI values (r = -0.19, R2 = 0.037 with p = 0.051). Similarly, the proposed RCM U-Net had noticeably better r and R2 values (r = 0.25, R2 = 0.065 with p = 0.007) for the correlation with the renal percentage fibrosis than the Manual ROI results (r = 0.3, R2 = 0.091 and p = 0.0013). Using a linear mixed model, T1 was significantly higher in the medulla than in the cortex (p<0.0001) and significantly lower in patients with cellular rejection when compared to both patients without rejection and those with humoral rejection (p<0.001). There was no significant difference in T1 between patients with and without humoral rejection (p = 0.43), nor between the types of T1 measurements (Gold standard manual versus automated RCM U-Net) (p = 0.7). The cortico-medullary area ratio measured by the RCM U-Net was significantly increased in case of cellular rejection by comparison to humoral rejection (1.6 +/- 0.39 versus 0.99 +/- 0.32, p = 0.019). In conclusion, the proposed RCM U-Net provides more robust auto-segmented cortex and medulla than the other standard CNNs allowing a good correlation of ΔT1 with eGFR and fibrosis as reported in literature as well as the differentiation of cellular and humoral transplant rejection. Therefore, the proposed approach is a promising alternative to the gold standard manual ROI method to measure T1 values without user interaction, which helps to reduce analysis time and improves reproducibility.

show abstract

Section: Plos Onementioning

confidence: 99%

“…cortex, medulla etc. Several studies [8][9][10][11][12][13] have concentrated on manual and automatic segmentation of kidneys for specific renal cysts or total kidney volume using DL techniques.…”

Section: Introductionmentioning

confidence: 99%

Validation of automatically measured T1 map cortico-medullary difference (ΔT1) for eGFR and fibrosis assessment in allograft kidneys

et al. 2023

View full text Add to dashboard Cite

show abstract

“…The percent difference in the cortex is 3.9478. Quantitative results show that our workflow can serve as the state-of-the-art volumetric measurement compared to prior kidney characterization pipeline [23].…”

Section: Characterization Of Renal Structuresmentioning

confidence: 99%

Characterizing Renal Structures with 3D Block Aggregate Transformers

Yu¹,

Tang²,

Zhou³

et al. 2022

Preprint

Self Cite

View full text Add to dashboard Cite

Efficiently quantifying renal structures can provide distinct spatial context and facilitate biomarker discovery for kidney morphology. However, the development and evaluation of transformer model to segment the renal cortex, medulla, and collecting system remains challenging due to data inefficiency. Inspired by the hierarchical structures in vision transformer, we propose a novel method using 3D block aggregation transformer for segmenting kidney components on contrast-enhanced CT scans. We construct the first cohort of renal substructures segmentation dataset with 116 subjects under institutional review board (IRB) approval. Our method yields the state-of-the-art performance (Dice of 0.8467) against the baseline approach of 0.8308 with the data-efficient design. The Pearson R achieves 0.9891 between the proposed method and manual standards, and indicates the strong correlation and reproducibility for volumetric analysis. We extend the proposed method to the public KiTS dataset, the method leads to improved accuracy compared to transformer-based approaches. We show that the 3D block aggregation transformer can achieve local communication between sequence representations without modifying self-attention, and it can serve as an accurate and efficient quantification tool for characterizing renal structures.

show abstract

“…Tang et al took a similar approach where early and late arterial phase scans were used to train a patch-based network to segment renal structures. 5 The group found the model to perform adequately on test data, with no distinction between late and early arterial phase performance. Lee et al attempted to reduce the dependency of labeling multiple phases by using paired samples where only the contrast-enhanced volume was annotated.…”

mentioning

confidence: 99%

Comparative analysis of multiphase CT volumetric kidney segmentation: fine-tuning to domain adaptation

Correa,

Fathizadeh,

Al Khafaji

et al. 2024

Medical Imaging 2024: Computer-Aided Diagnosis

View full text Add to dashboard Cite

Development of kidney segmentation models has largely focused on using contrast-enhanced CT exams. The KiTS segmentation challenge, in particular, has provided a benchmark using 300 annotated arterial phase CT scans. 1 Review of the best performing models identifies 3D Unet models with residual connections as the best performing models for kidney segmentation. 2 Li et al. found a U-net architecture with residual connections to provide the best performance for their segmentation task. 3 Their work focused on segmenting kidney parenchyma alongside kidney stones using a dataset of 257 studies that was recently made available. Yu et al. investigated the ability to train a multi-organ nn-Unet model using simultaneous contrast and non-contrast images. The authors found the model to achieve high dice scores for kidney segmentation with average dice scores of 0.96. 4 Inspection of output segmentation found the model to underperform for non-contrast images as measured by smaller quality assessment scores. Tang et al. took a similar approach where early and late arterial phase scans were used to train a patch-based network to segment renal structures. 5 The group found the model to perform adequately on test data, with no distinction between late and early arterial phase performance. Lee et al. attempted to reduce the dependency of labeling multiple phases by using paired samples where only the contrast-enhanced volume was annotated. 6 They were able to outperform existing models but were limited by the need for correct anatomical correspondence between scans. Ananda et al. removed the dependency for paired samples by training a dual discriminator-based network where the model is trained in three phases. 7 One phase ensures the consistency of segmentation of contrast phase images; the following two phases ensure the image encoding and output maps are not significantly different from contrast and non-contrast images. Dinsdale et al. implemented a similar multi-step approach to improve segmentation quality by improving resiliency to agerelated physiological changes in Brain MRI, where cross entropy(CE) loss is utilized for training a discriminator on the patient's age using features from the bottleneck and segmentation map. In contrast, the final phase uses a confusion loss, penalizing the model for having greater confidence for a particular age. 8 Despite all the modeling effort, the application of evaluating patients with impaired kidney function is challenging since variation occurs not only due to phase of contrast but also the level of renal function, and thus the consistency of imaging appearance is not guaranteed. Establishing the need for techniques that would improve the robustness of kidney segmentation models. Within the scope of this work, we proposed various techniques to generate models resilient to different contrast phases and externally validated the models. METHODS Datasetsi. Contrast-enhanced CT exams and kidney segmentation masks -The publicly accessible KiTS21 dataset has 300 arterial phase de-identified...

show abstract

Renal cortex, medulla and pelvicaliceal system segmentation on arterial phase CT images with random patch-based networks

Cited by 6 publications

References 21 publications

Validation of automatically measured T1 map cortico-medullary difference (ΔT1) for eGFR and fibrosis assessment in allograft kidneys

Validation of automatically measured T1 map cortico-medullary difference (ΔT1) for eGFR and fibrosis assessment in allograft kidneys

Characterizing Renal Structures with 3D Block Aggregate Transformers

Comparative analysis of multiphase CT volumetric kidney segmentation: fine-tuning to domain adaptation

Contact Info

Product

Resources

About