Unsupervised learning of depth estimation from imperfect rectified stereo laparoscopic images

BACKGROUND

Learning-based methods have achieved remarkable performances on depth estimation. However, the premise of most self-learning and unsupervised learning methods is built on rigorous, geometrically-aligned stereo rectification. The performances of these methods degrade when the rectification is not accurate. Therefore, we explore an approach for unsupervised depth estimation from stereo images that can handle imperfect camera parameters.

METHODS

We propose an unsupervised deep convolutional network that takes rectified stereo image pairs as input and outputs corresponding dense disparity maps. First, a new vertical correction module is designed for predicting a correction map to compensate for the imperfect geometry alignment. Second, the left and right images, which are reconstructed based on the input image pair and corresponding disparities as well as the vertical correction maps, are regarded as the outputs of the generative term of the generative adversarial network (GAN). Then, the discriminator term of the GAN is used to distinguish the reconstructed images from the original inputs to force the generator to output increasingly realistic images. In addition, a residual mask is introduced to exclude pixels that conflict with the appearance of the original image in the loss calculation.

RESULTS

The proposed model is validated on the publicly available Stereo Correspondence and Reconstruction of Endoscopic Data (SCARED) dataset and the average MAE is 3.054 mm.

CONCLUSION

Our model can effectively handle imperfect rectified stereo images for depth estimation.

A study of generalization and compatibility performance of 3D U-Net segmentation on multiple heterogeneous liver CT datasets

BACKGROUND

Most existing algorithms have been focused on the segmentation from several public Liver CT datasets scanned regularly (no pneumoperitoneum and horizontal supine position). This study primarily segmented datasets with unconventional liver shapes and intensities deduced by contrast phases, irregular scanning conditions, different scanning objects of pigs and patients with large pathological tumors, which formed the multiple heterogeneity of datasets used in this study.

METHODS

The multiple heterogeneous datasets used in this paper includes: (1) One public contrast-enhanced CT dataset and one public non-contrast CT dataset; (2) A contrast-enhanced dataset that has abnormal liver shape with very long left liver lobes and large-sized liver tumors with abnormal presets deduced by microvascular invasion; (3) One artificial pneumoperitoneum dataset under the pneumoperitoneum and three scanning profiles (horizontal/left/right recumbent position); (4) Two porcine datasets of Bama type and domestic type that contains pneumoperitoneum cases but with large anatomy discrepancy with humans. The study aimed to investigate the segmentation performances of 3D U-Net in: (1) generalization ability between multiple heterogeneous datasets by cross-testing experiments; (2) the compatibility when hybrid training all datasets in different sampling and encoder layer sharing schema. We further investigated the compatibility of encoder level by setting separate level for each dataset (i.e., dataset-wise convolutions) while sharing the decoder.

RESULTS

Model trained on different datasets has different segmentation performance. The prediction accuracy between LiTS dataset and Zhujiang dataset was about 0.955 and 0.958 which shows their good generalization ability due to that they were all contrast-enhanced clinical patient datasets scanned regularly. For the datasets scanned under pneumoperitoneum, their corresponding datasets scanned without pneumoperitoneum showed good generalization ability. Dataset-wise convolution module in high-level can improve the dataset unbalance problem. The experimental results will facilitate researchers making solutions when segmenting those special datasets.

CONCLUSIONS

(1) Regularly scanned datasets is well generalized to irregularly ones. (2) The hybrid training is beneficial but the dataset imbalance problem always exits due to the multi-domain homogeneity. The higher levels encoded more domain specific information than lower levels and thus were less compatible in terms of our datasets.

A radiomics-based nomogram for preoperative T staging prediction of rectal cancer

Purpose

To investigate the value of a radiomics-based nomogram in predicting preoperative T staging of rectal cancer.

Methods

A total of 268 eligible rectal cancer patients from August 2012 to December 2018 were enrolled and allocated into two datasets: training (n = 188) and validation datasets (n = 80). Another set of 32 patients from January 2019 to July 2019 was included in a prospective analysis. Pretreatment T2-weighted images were used to radiomics features extraction. Feature selection and radiomics score (Rad-score) construction were performed through a least absolute shrinkage and selection operator regression analysis. The nomogram, which included Rad-scores and clinical factors, was built using multivariate logistic regression. Discrimination, calibration, and clinical utility were used to evaluate the performance of the nomogram.

Results

The Rad-score containing nine selected features was significantly related to T staging. Patients who had locally advanced rectal cancer (LARC) generally had higher Rad-scores than patients with early-stage rectal cancer. The nomogram incorporated Rad-scores and carcinoembryonic antigen levels and showed good discrimination, with an area under the curve (AUC) of 0.882 (95% confidence interval [CI] 0.835–0.930) in the training dataset and 0.846 (95% CI 0.757–0.936) in the validation dataset. The calibration curves confirmed high goodness of fit, and the decision curve analysis revealed the clinical value. A prospective analysis demonstrated that the AUC of the nomogram to predict LARC was 0.859 (95% CI 0.730–0.987).

Conclusion

A radiomics-based nomogram is a novel method for predicting LARC and can provide support in clinical decision making.

Supplementary Information

The online version contains supplementary material available at 10.1007/s00261-021-03137-1.

Morphologic Change of In Vivo Porcine Liver Under 13 mm Hg Pneumoperitoneum Pressure

BACKGROUND

Clinically, the total and residual liver volume must be accurately calculated before major hepatectomy. However, liver volume might be influenced by pneumoperitoneum during surgery. Changes in liver volume change also affect the accuracy of simulation and augmented reality navigation systems, which are commonly first validated in animal models. In this study, the morphologic changes in porcine livers in vivo under 13 mm Hg pneumoperitoneum pressure were investigated.

MATERIALS AND METHODS

Twenty male pigs were scanned with contrast-enhanced computed tomography without pneumoperitoneum and with 13 mm Hg pneumoperitoneum pressure.

RESULTS

The surface area and volume of the liver and the vascular diameter of the aortic lumen, inferior vena cava lumen, and portal vein lumen were measured. There were statistically significant differences in the surface area and volume of the liver (P=0.000), transverse diameter of the portal vein (P=0.038), longitudinal diameter of the inferior vena cava (P=0.033), longitudinal diameter of the portal vein (P=0.036), vascular cross-sectional area of the inferior vena cava (P=0.028), and portal vein (P=0.038) before and after 13 mm Hg pneumoperitoneum pressure.

CONCLUSIONS

This study indicated that the creation of pneumoperitoneum at 13 mm Hg pressure in a porcine causes liver morphologic alterations affecting the area and volume, as well as the diameter of a blood vessel.

Against spatial-temporal discrepancy: contrastive learning-based network for surgical workflow recognition

PURPOSE

Automatic workflow recognition from surgical videos is fundamental and significant for developing context-aware systems in modern operating rooms. Although many approaches have been proposed to tackle challenges in this complex task, there are still many problems such as the fine-grained characteristics and spatial-temporal discrepancies in surgical videos.

METHODS

We propose a contrastive learning-based convolutional recurrent network with multi-level prediction to tackle these problems. Specifically, split-attention blocks are employed to extract spatial features. Through a mapping function in the step-phase branch, the current workflow can be predicted on two mutual-boosting levels. Furthermore, a contrastive branch is introduced to learn the spatial-temporal features that eliminate irrelevant changes in the environment.

RESULTS

We evaluate our method on the Cataract-101 dataset. The results show that our method achieves an accuracy of 96.37% with only surgical step labels, which outperforms other state-of-the-art approaches.

CONCLUSION

The proposed convolutional recurrent network based on step-phase prediction and contrastive learning can leverage fine-grained characteristics and alleviate spatial-temporal discrepancies to improve the performance of surgical workflow recognition.

Augmented Reality Navigation for Stereoscopic Laparoscopic Anatomical Hepatectomy of Primary Liver Cancer: Preliminary Experience

Background

Accurate determination of intrahepatic anatomy remains challenging for laparoscopic anatomical hepatectomy (LAH). Laparoscopic augmented reality navigation (LARN) is expected to facilitate LAH of primary liver cancer (PLC) by identifying the exact location of tumors and vessels. The study was to evaluate the safety and effectiveness of our independently developed LARN system in LAH of PLC.

Methods

From May 2018 to July 2020, the study included 85 PLC patients who underwent three-dimensional (3D) LAH. According to whether LARN was performed during the operation, the patients were divided into the intraoperative navigation (IN) group and the non-intraoperative navigation (NIN) group. We compared the preoperative data, perioperative results and postoperative complications between the two groups, and introduced our preliminary experience of this novel technology in LAH.

Results

There were 44 and 41 PLC patients in the IN group and the NIN group, respectively. No significant differences were found in preoperative characteristics and any of the resection-related complications between the two groups (All P > 0.05). Compared with the NIN group, the IN group had significantly less operative bleeding (P = 0.002), lower delta Hb% (P = 0.039), lower blood transfusion rate (P < 0.001), and reduced postoperative hospital stay (P = 0.003). For the IN group, the successful fusion of simulated surgical planning and operative scene helped to determine the extent of resection.

Conclusions

The LARN contributed to the identification of important anatomical structures during LAH of PLC. It reduced vascular injury and accelerated postoperative recovery, showing a potential application prospects in liver surgery.

MDCC-Net: Multiscale double-channel convolution U-Net framework for colorectal tumor segmentation

PURPOSE

Multiscale feature fusion is a feasible method to improve tumor segmentation accuracy. However, current multiscale networks have two common problems: 1. Some networks only allow feature fusion between encoders and decoders of the same scale. It is obvious that such feature fusion is not sufficient. 2. Some networks have too many dense skip connections and too much nesting between the coding layer and the decoding layer, which causes some features to be lost and means that not enough information will be learned from multiple scales. To overcome these two problems, we propose a multiscale double-channel convolution U-Net (MDCC-Net) framework for colorectal tumor segmentation.

METHODS

In the coding layer, we designed a dual-channel separation and convolution module and then added residual connections to perform multiscale feature fusion on the input image and the feature map after dual-channel separation and convolution. By fusing features at different scales in the same coding layer, the network can fully extract the detailed information of the original image and learn more tumor boundary information.

RESULTS

The segmentation results show that our proposed method has a high accuracy, with a Dice similarity coefficient (DSC) of 83.57%, which is an improvement of 9.59%, 6.42%, and 1.57% compared with nnU-Net, U-Net, and U-Net++, respectively.

CONCLUSION

The experimental results show that our proposed method has good performance in the segmentation of colorectal tumors and is close to the expert level. The proposed method has potential clinical applicability.

Radiomic Feature-Based Predictive Model for Microvascular Invasion in Patients With Hepatocellular Carcinoma

Objective

This study aimed to build and evaluate a radiomics feature-based model for the preoperative prediction of microvascular invasion (MVI) in patients with hepatocellular carcinoma.

Methods

A total of 145 patients were retrospectively included in the study pool, and the patients were divided randomly into two independent cohorts with a ratio of 7:3 (training cohort: n = 101, validation cohort: n = 44). For a pilot study of this predictive model another 18 patients were recruited into this study. A total of 1,231 computed tomography (CT) image features of the liver parenchyma without tumors were extracted from portal-phase CT images. A least absolute shrinkage and selection operator (LASSO) logistic regression was applied to build a radiomics score (Rad-score) model. Afterwards, a nomogram, including Rad-score as well as other clinicopathological risk factors, was established with a multivariate logistic regression model. The discrimination efficacy, calibration efficacy, and clinical utility value of the nomogram were evaluated.

Results

The Rad-score scoring model could predict MVI with the area under the curve (AUC) of 0.637 (95% CI, 0.516–0.758) in the training cohort as well as of 0.583 (95% CI, 0.395–0.770) in the validation cohort; however, the aforementioned discriminative approach could not completely outperform those existing predictors (alpha fetoprotein, neutrophilic granulocyte, and preoperative hemoglobin). The individual predictive nomogram which included the Rad-score, alpha fetoprotein, neutrophilic granulocyte, and preoperative hemoglobin showed a better discrimination efficacy with AUC of 0.865 (95% CI, 0.786–0.944), which was higher than the conventional methods’ AUCs (nomogram vs Rad-score, alpha fetoprotein, neutrophilic granulocyte, and preoperative hemoglobin at P < 0.001, P = 0.025, P < 0.001, and P = 0.001, respectively). When applied to the validation cohort, the nomogram discrimination efficacy was still outbalanced those above mentioned three remaining methods (AUC: 0.705; 95% CI, 0.537–0.874). The calibration curves of this proposed method showed a satisfying consistency in both cohorts. A prospective pilot analysis showed that the nomogram could predict MVI with an AUC of 0.844 (95% CI, 0.628–1.000).

Conclusions

The radiomics feature-based predictive model improved the preoperative prediction of MVI in HCC patients significantly. It could be a potentially valuable clinical utility.

ILK participates in renal interstitial fibrosis by altering the phenotype of renal tubular epithelial cells via TGF-β1/smad pathway

OBJECTIVE

To explore the specific role of ILK (integrin-linked kinase) in regulating renal fibrosis and its underlying mechanism.

MATERIALS AND METHODS

NRK-52E cells were induced by transforming growth factor-β1 (TGF-β1) for observing phenotype change. Renal tubular epithelial cell marker, fibrosis marker and expression level of ILK in NRK-52E cells were also detected. After overexpression of ILK, phenotype change of NRK-52E cells was observed. For in vivo experiments, we constructed UUO (unilateral ureteral obstruction) model in CD1 mice. Renal tubular epithelial cell marker, fibrosis marker and expression level of ILK in UUO mice were detected. The regulatory effect of ILK on renal fibrosis was detected after injection of ILK overexpression plasmid. Western blot was performed to detect related genes in the TGF-β1/smad pathway.

RESULTS

Accompanied by the TGF-β1-induced phenotype change in NRK-52E cells, both mRNA and protein levels of ILK were upregulated. Overexpression of ILK remarkably stimulated the phenotype change in NRK-52E cells. Similarly, ILK was highly expressed in UUO mice. Renal fibrosis was aggravated after injection of ILK overexpression plasmid in UUO mice. Western blot results showed that expressions of p-smad3 and smad3 were upregulated during the process of renal fibrosis.

CONCLUSIONS

ILK is upregulated during the process of renal fibrosis. ILK participates in the development of renal fibrosis by altering phenotypes of renal tubular epithelial cells via a TGF-β1/smad pathway.

Augmented reality navigation for liver resection with a stereoscopic laparoscope

OBJECTIVE

Understanding the three-dimensional (3D) spatial position and orientation of vessels and tumor(s) is vital in laparoscopic liver resection procedures. Augmented reality (AR) techniques can help surgeons see the patient's internal anatomy in conjunction with laparoscopic video images.

METHOD

In this paper, we present an AR-assisted navigation system for liver resection based on a rigid stereoscopic laparoscope. The stereo image pairs from the laparoscope are used by an unsupervised convolutional network (CNN) framework to estimate depth and generate an intraoperative 3D liver surface. Meanwhile, 3D models of the patient's surgical field are segmented from preoperative CT images using V-Net architecture for volumetric image data in an end-to-end predictive style. A globally optimal iterative closest point (Go-ICP) algorithm is adopted to register the pre- and intraoperative models into a unified coordinate space; then, the preoperative 3D models are superimposed on the live laparoscopic images to provide the surgeon with detailed information about the subsurface of the patient's anatomy, including tumors, their resection margins and vessels.

RESULTS

The proposed navigation system is tested on four laboratory ex vivo porcine livers and five operating theatre in vivo porcine experiments to validate its accuracy. The ex vivo and in vivo reprojection errors (RPE) are 6.04 ± 1.85 mm and 8.73 ± 2.43 mm, respectively.

CONCLUSION AND SIGNIFICANCE

Both the qualitative and quantitative results indicate that our AR-assisted navigation system shows promise and has the potential to be highly useful in clinical practice.

Elevated aerobic glycolysis in renal tubular epithelial cells influences the proliferation and differentiation of podocytes and promotes renal interstitial fibrosis

OBJECTIVE

The aim of this study was to elaborate the influence of changing energy metabolism pattern of renal tubular epithelial cells in the process of renal interstitial fibrosis on podocytes. Meanwhile, we also investigated the relationship between energy metabolism pattern and the development of renal interstitial fibrosis.

MATERIALS AND METHODS

We established a model of renal interstitial fibrosis by unilateral ureteral obstruction (UUO). The protein and messenger RNA (mRNA) expression of fibrosis signs, such as α-smooth muscle actin (α-SMA) and fibronectin (FN) were detected. We also measured the protein and mRNA expression of key glycolytic enzymes, including pyruvate kinase muscle isozyme 2 (PKM2) and human glandular kallikrein 2 (HK2). The proliferation and differentiation of podocytes during fibrosis were observed by monitoring the expression of nephrin and myocardin. In vitro experiments, primary podocytes were extracted, cultured, and stimulated with lactate. Then the alterations during the process were observed. Finally, PKM2 expression was inhibited by intravenous infusion of the plasmid. The link between the expression of marker protein as well as differentiation protein in podocytes and renal interstitial fibrosis was analyzed.

RESULTS

During the process of renal interstitial fibrosis, phenotypic changes and enhanced expression of fibrosis and proliferation markers were found in fibroblasts. Meanwhile, in renal tubular epithelial cells, increased expression of key enzymes of glycolysis, the level of glycolysis as well as lactate metabolites cooperatively led to hypoxic and acidic environment, eventually inhibiting the proliferation and differentiation of podocytes and aggravating fibrosis. When the level of glycolysis in renal tubular epithelial cells was reduced, the number and function of podocytes were partially restored, and renal interstitial fibrosis was alleviated.

CONCLUSIONS

During renal interstitial fibrosis, glycolysis of renal tubular epithelial cell was increased, leading to the recodification of energy metabolism. This process affected the number and function of podocytes and aggravated renal interstitial fibrosis.

Details preserved unsupervised depth estimation by fusing traditional stereo knowledge from laparoscopic images

Depth estimation plays an important role in vision-based laparoscope surgical navigation systems. Most learning-based depth estimation methods require ground truth depth or disparity images for training; however, these data are difficult to obtain in laparoscopy. The authors present an unsupervised learning depth estimation approach by fusing traditional stereo knowledge. The traditional stereo method is used to generate proxy disparity labels, in which unreliable depth measurements are removed via a confidence measure to improve stereo accuracy. The disparity images are generated by training a dual encoder-decoder convolutional neural network from rectified stereo images coupled with proxy labels generated by the traditional stereo method. A principled mask is computed to exclude the pixels, which are not seen in one of views due to parallax effects from the calculation of loss function. Moreover, the neighbourhood smoothness term is employed to constrain neighbouring pixels with similar appearances to generate a smooth depth surface. This approach can make the depth of the projected point cloud closer to the real surgical site and preserve realistic details. The authors demonstrate the performance of the method by training and evaluation with a partial nephrectomy da Vinci surgery dataset and heart phantom data from the Hamlyn Centre.

Delineating functional segregations of the human middle temporal gyrus with resting-state functional connectivity and coactivation patterns

Although the middle temporal gyrus (MTG) has been parcellated into subregions with distinguished anatomical connectivity patterns, whether the structural topography of MTG can inform functional segregations of this area remains largely unknown. Accumulating evidence suggests that the brain's underlying organization and function can be directly and effectively delineated with resting‐state functional connectivity (RSFC) by identifying putative functional boundaries between cortical areas. Here, RSFC profiles were used to explore functional segregations of the MTG and defined four subregions from anterior to posterior in two independent datasets, which showed a similar pattern with MTG parcellation scheme obtained using anatomical connectivity. The functional segregations of MTG were further supported by whole brain RSFC, coactivation, and specific RFSC, and coactivation mapping. Furthermore, the fingerprint with predefined 10 networks and functional characterizations of each subregion using meta‐analysis also identified functional distinction between subregions. The specific connectivity analysis and functional characterization indicated that the bilateral most anterior subregions mainly participated in social cognition and semantic processing; the ventral middle subregions were involved in social cognition in left hemisphere and auditory processing in right hemisphere; the bilateral ventro‐posterior subregions participated in action observation, whereas the left subregion was also involved in semantic processing; both of the dorsal subregions in superior temporal sulcus were involved in language, social cognition, and auditory processing. Taken together, our findings demonstrated MTG sharing similar structural and functional topographies and provide more detailed information about the functional organization of the MTG, which may facilitate future clinical and cognitive research on this area.

Retraction: Randomized controlled trial of vitamin D supplementation in children with autism spectrum disorder

The above article, published in print in the Jan 2018 issue of the Journal of Child Psychology & Psychiatry and online in Wiley Online Library (wileyonlinelibrary.com), has been retracted by the JCPP Editor-in-Chief, Edmund Sonuga-Barke, and John Wiley & Sons. Following a series of communications from readers highlighting concerns about the paper (now published on the journal website), the journal editors requested that the authors send them the raw data from the trial. In response the authors informed the editors that; (i) the electronic data base had been lost following a computer outage and (ii) that they could send only 95 out of 120 hard-copy participant data sheets as one site had closed and was no longer contactable. The substantial data loss in and of itself posed a serious difficulty in verifying the correctness of the data presented in the paper. The JCPP then analysed the data from the 95 cases itself. A number of significant discrepancies emerged between the re-analysis and the findings reported in the paper both in terms of means and standard deviations of key outcome variables across the trial. These involved very substantial differences that we judged to be extremely unlikely to have arisen due to variations in composition of the original and re-analysed samples. We also discovered previously unidentified/reported problems with missing data and recording irregularities regarding changes in treatment regimen and subject identifiers. As a result of these issues the Editors no longer have confidence in the findings reported in the original paper. Based on all these matters combined and following published guidance from the Committee on Publishing Ethics (COPE) and Wiley's Best Practice Guidelines on Publishing Ethics, we have decided that the only course of action available to us is to retract the paper.

Unsupervised binocular depth prediction network for laparoscopic surgery

Minimally invasive surgery (MIS) is characterized by less trauma, shorter recovery time, and lower postoperative infection rate. The two-dimensional (2D) laparoscopic imaging lacks depth perception and does not provide quantitative depth information, thereby limiting precise and complex surgical operations. Three-dimensional (3D) laparoscopic imaging provides surgeons depth perception. This study aims to 3D reconstruction of the surgical scene based on the disparity map generated by the depth estimation algorithm. An unsupervised learning autoencoder method was proposed to calculate the accurate disparity with a 101-layer residual convolutional network. The loss function included three parts: left-right consistency loss, structure similarity loss, and reconstruction error loss, the combination can improve reconstruction accuracy and robustness. The method was validated on a Hamlyn Center Laparoscopic/Endoscopic Video Dataset. The structural similarity index (SSIM) is 0.8349 ± 0.0523 and the peak signal-to-noise ratio (PSNR) is 14.4957 ± 1.9676. The depth prediction network has high accuracy and robustness. The average time to produce each disparity map is about 16 ms. The experimental result shows that the proposed depth estimation method can offer dense disparity map, and can meet surgical real-time requirement. Future work will focus on network structure optimization and loss function design, transfer learning to improve the robustness and accuracy further.

Unsupervised binocular depth prediction network for laparoscopic surgery

Minimally invasive laparoscopic surgery is associated with small wounds and short recovery time, reducing postoperative infections. Traditional two-dimensional (2D) laparoscopic imaging lacks depth perception and does not provide quantitative depth information, thereby limiting the field of vision and operation during surgery. However, three-dimensional (3D) laparoscopic imaging from 2 D images lets surgeons have a depth perception. However, the depth information is not quantitative and cannot be used for robotic surgery. Therefore, this study aimed to reconstruct the accurate depth map for binocular 3 D laparoscopy. In this study, an unsupervised learning method was proposed to calculate the accurate depth while the ground-truth depth was not available. Experimental results proved that the method not only generated accurate depth maps but also provided real-time computation, and it could be used in minimally invasive robotic surgery.

A radiomics-based formula for the preoperative prediction of postoperative pancreatic fistula in patients with pancreaticoduodenectomy

Objective

The objective of the study was to develop and validate a radiomics-based formula for the preoperative prediction of postoperative pancreatic fistula (POPF) in patients undergoing pancreaticoduodenectomy (PD).

Materials and methods

A total of 117 consecutive patients who underwent PD were enrolled in this retrospective study. Radiomics features were extracted from portal venous phase computed tomography of the above patients. The least absolute shrinkage and selection operator logistic regression was used to construct a formula of Rad-score calculation. Then the performance of the formula was assessed with standard pancreatic Fistula Risk Score.

Results

The Rad-score could predict POPF with an area under the curve (AUC) of 0.8248 in the training cohort and of 0.7609 in the validation cohort. Patients who had experienced POPF generally had a statistically higher Rad-score than those who had not experienced POPF in both cohorts. The AUC of the Rad-score was statistically higher than the Fistula Risk Score for predicting POPF in both the training and validation cohort.

Conclusion

A novel radiomics-based formula was developed and validated for predicting POPF in patients who underwent PD, which provides a new method for identifying POPF risks and may help to improve informed decision-making in the prevention of POPF at low cost.

A radiomics-based nomogram for the preoperative prediction of posthepatectomy liver failure in patients with hepatocellular carcinoma

OBJECTIVES

To develop and validate a radiomics-based nomogram for the preoperative prediction of posthepatectomy liver failure (PHLF) in patients with hepatocellular carcinoma (HCC).

METHODS

One hundred twelve consecutive HCC patients who underwent hepatectomy were included in the study pool (training cohort: n = 80, validation cohort: n = 32), and another 13 patients were included in a pilot prospective analysis. A total of 713 radiomics features were extracted from portal-phase computed tomography (CT) images. A logistic regression was used to construct a radiomics score (Rad-score). Then a nomogram, including Rad-score and other risk factors, was built with a multivariate logistic regression model. The discrimination, calibration and clinical utility of nomogram were evaluated.

RESULTS

The Rad-score could predict PHLF with an AUC of 0.822 (95% CI, 0.726-0.917) in the training cohort and of 0.762 (95% CI, 0.576-0.948) in the validation cohort; however, the approach could not completely outmatch the existing methods (CP [Child-Pugh], MELD [Model of End Stage Liver Disease], ALBI [albumin-bilirubin]). The individual predictive nomogram that included the Rad-score, MELD and performance status (PS) showed better discrimination with an AUC of 0.864 (95% CI, 0.786-0.942), which was higher than the AUCs of the conventional methods (nomogram vs CP, MELD, and ALBI at P < 0.001, P < 0.005, and P < 0.005, respectively). In the validation cohort, the nomogram discrimination was also superior to those of the other three methods (AUC: 0.896; 95% CI, 0.774-1.000). The calibration curves showed good agreement in both cohorts, and the decision curve analysis of the entire cohort revealed that the nomogram was clinically useful. A pilot prospective analysis showed that the radiomics nomogram could predict PHLF with an AUC of 0.833 (95% CI, 0.591-1.000).

CONCLUSIONS

A nomogram based on the Rad-score, MELD, and PS can predict PHLF.