Application of Comprehensive Pelvic Floor Rehabilitation Therapy in Congestion Syndrome with Pelvic Oblique: A Case Report.

Background: Pelvic congestion syndrome (PCS), also known as ovarian vein syndrome, is one of the key causes of chronic pelvic pain. The present study combined pelvic floor myofascial manipulation, uterine conditioning, and improved Kegel exercise for the treatment of a PCS patient with pelvic inclination to provide a reference for clinical therapy. Case description: A 29-year-old female was admitted to our hospital on 20th April 2023, with the main complaint being repeated lower abdominal and lumbosacral pain with frequent urination (urine volume < 200 ml/time), external genital itching accompanied by increased secretion for more than 5 years. The patient was treated with pelvic floor myofascial manipulation, uterine conditioning, and improved Kegel exercise 6 times. Pelvic magnetic resonance imaging (MRI) examination, pelvic X-ray examination, overall posture assessment, and related functional status were observed and evaluated. Conclusions: After comprehensive pelvic floor rehabilitation treatment containing pelvic floor myofascial manipulation, uterine conditioning, and improved Kegel training, the symptoms and signs of the patient were significantly improved, and the effect was obvious.

Demethylnobiletin ameliorates cerebral ischemia-reperfusion injury in rats through Nrf2/HO-1 signaling pathway.

BACKGROUND: Demethylnobiletin (DN), with a variety of biological activities, is a polymethoxy-flavanone (PMF) found in citrus. In the present study, we explored the biological activities and potential mechanism of DN to improve cerebral ischemia reperfusion injury (CIRI) in rats, and identified DN as a novel neuroprotective agent for patients with ischemic brain injury. METHODS: Rat CIRI models were established via middle cerebral artery occlusion (MCAO). Primary nerve cells were isolated and cultured in fetal rat cerebral cortex in vitro, and oxygen-glucose deprivation/reperfusion (OGD/R) models of primary nerve cells were induced. After intervention with DN with different concentrations in MCAO rats and OGD/R nerve cells, 2,3,5-triphenyltetrazolium chloride staining was used to quantify cerebral infarction size in CIRI rats. Modified neurological severity score was utilized to assess neurological performance. Histopathologic staining and live/dead cell-viability staining was used to observe apoptosis. Levels of glutathione (GSH), superoxide dismutase (SOD), reactive oxygen species (ROS) and malondialdehyde (MDA) in tissues and cells were detected using commercial kits. DN level in serum and cerebrospinal fluid of MCAO rats were measured by liquid chromatography tandem mass spectrometry. In addition, expression levels of proteins like Kelch like ECH associated protein 1 (Keap1), nuclear factor erythroid 2-related factor 2 (Nfr2) and heme oxygenase 1 (HO-1) in the Nrf2/HO-1 pathway, and apoptosis-related proteins like Cleaved caspase-3, BCL-2-associated X protein (Bax) and B-cell lymphoma-2 (Bcl-2) were determined by Western blot and immunofluorescence. RESULTS: DN can significantly enhance neurological function recovery by reducing cerebral infarction size and weakening neurocytes apoptosis in MCAO rats. It was further found that DN could improve oxidative stress (OS) injury of nerve cells by bringing down MDA and ROS levels and increasing SOD and GSH levels. Notably, DN exerts its pharmacological influences through entering blood-brain barrier. Mechanically, DN can reduce Keap1 expression while activate Nrf2 and HO-1 expression in neurocytes. CONCLUSIONS: The protective effect of DN on neurocytes have been demonstrated in both in vitro and in vivo circumstances. It deserves to be developed as a potential neuroprotective agent through regulating the Nrf2/HO-1 signaling pathway to ameliorate neurocytes impairment caused by OS.

Regulating the thickness of nanofiltration membranes for efficient water purification.

Fabrication of an organic polymer nanofiltration membrane with both high water permeability and high salt rejection is still a big challenge. Herein, phytic acid (PhA)-modified graphene oxide (GO) was used as the membrane thickness modifier, which was introduced into the thin-film nanoparticle composite (TFN) membrane via in situ interfacial polymerization (IP) on a porous substrate. The water flux of the optimally tuned TFN-GP-0.2 composite membrane is 48.9 L m-2 h-1, which is 1.3 times that of the pristine thin-film composite (TFC) nanofiltration membrane (37.9 L m-2 h-1) (GP represents the PhA modified GO composite). The rejection rate of 2000 ppm MgSO4 for TFN-GP-0.2 membranes was maintained at 97.5%. The increased water flux of the TFN-GP composite membrane compared to that of the TFN nanofiltration membrane was mainly attributed to enhanced hydrophilicity and reduced thickness of the polyamide (PA) layer. Molecular dynamics (MD) simulations confirm that the diffusion rate of amine monomers is reduced by the presence of a GP complex in the IP process, which facilitates the formation of PA layer with thinner thickness. In addition, the TFN-GP-0.2 composite membrane also showed good long-term stability; after 12 h of continuous operation, the water flux only decreased by 0.1%. This study sheds new light on the development of GO-based nanofiltration for potential implementation, as well as a unique concept for manufacturing high-performance nanofiltration membranes.

Effects of Early Weight-Bearing Treadmill Training Combined with Pre-Emptive Analgesia on Femoral Fracture Recovery.

Background: The effect of pre-emptive analgesia plus early weight-bearing treadmill training (EWBTT) on healing and motor function recovery of femoral shaft fracture is not clear. Methods: A total of 60 SD male rats were randomly allocated into 4 groups: group A (pre-emptive analgesia with EWBTT), group B (pre-emptive analgesia with delayed weight-bearing treadmill training, DWBTT), group C (pre-emptive analgesia with no weight-bearing), and group D (EWBTT with no pre-emptive analgesia). All rats were molded by internal fixation with Kirschner wire after right femoral shaft fracture. In groups A, B, and C, tramadol was intramuscularly injected 15 minutes before surgery. EWBTT was performed at day 1 postoperatively in groups A and D, and DWBTT was performed at day 14 postoperatively in group B. Oblique plate test was accomplished to assess hindlimb motor function recovery of rats in each group. Status of fracture healing was assessed through digital radiography (DR). Hematoxylin-eosin (HE) staining and immunohistochemistry of bone morphogenetic protein-2 (MBP-2) and vascular endothelial growth factor (VEGF) in callus were performed to explore fracture healing. The expression of BMP-2 and VEGF protein in quadriceps femoris muscle was detected by Western blot technique and mRNA expression of BMP-2 and VEGF in callus ascertained via reverse transcription-polymerase chain reaction (RT-PCR) technique. Results: For oblique plate test, rats in group A outperformed those in groups B and C at all time points after operation. DR image revealed that large numbers of callus growth, blurred fracture line, and obvious continuous callus passing through the fracture line can be found in group A at day 28 postoperatively, which is the best healing status among all groups. HE staining of callus confirmed the optimal effect of healing for rats in group A. VEGF and BMP-2 expression by immunohistochemistry showed a significantly higher positive score for callus in group A while those in group C being the lowest at all time points postoperatively. Significantly higher expression level of VEGF and BMP-2 protein was detected in quadriceps femoris muscle from group A, which exceeded those in all other groups at all time points. RT-PCR testing proved the highest expression of BMP-2 and VEGF mRNA in callus of rats from group A, significantly higher than those of other groups. Conclusions: Both pre-emptive analgesia and EWBTT can effectively invoke the expression of VEGF and BMP-2 and promote recovery of hindlimb locomotor function in rats with femoral fracture, and the combination of them leads to more superior results.

Biomechanical effects of hybrid constructions in the treatment of noncontinuous cervical spondylopathy: a finite element analysis.

BACKGROUND: Hybrid construction (HC) may be an ideal surgical strategy than noncontinuous total disc replacement (TDR) and noncontinuous anterior cervical discectomy and fusion (ACDF) in the treatment of noncontinuous cervical spondylopathy. However, there is still no consensus on the segmental selection for ACDF or TDR in HC. The study aims to analyse the effects of different segment selection of TDR and ACDF on cervical biomechanical characteristics after HC surgery. METHODS: Twelve FEMs of C2-C7 were constructed based on CT images of 12 mild cervical spondylopathy volunteers. Two kinds of HC were introduced in our study: Fusion-arthroplasty group (Group 1), upper-level (C3/4) ACDF, and lower-level TDR (C5/6); Arthroplasty-fusion group (Group 2), upper-level (C3/4) TDR and lower-level ACDF (C5/6). The follow-load technique was simulated by applying an axial initial load of 73.6 N through the motion centre of FEM. A bending moment of 1.0 Nm was applied to the centre of C2 in all FEMs. Statistical analysis was carried out by SPSS 26.0. The significance threshold was 5% (P < 0.05). RESULTS: In the comparison of ROMs between Group 1 and Group 2, the ROM in extension (P = 0.016), and lateral bending (P = 0.038) of C4/5 were significantly higher in Group 1 group. The average intervertebral disc pressures at C2/3 in all directions were significantly higher in Group 1 than those in Group 2 (P < 0.005). The average contact forces in facet joints of C2/3 (P = 0.007) were significantly more than that in Group 2; however, the average contact forces in facet joints of C6/7 (P < 0.001) in Group 1 group were significantly less than that in Group 2. CONCLUSIONS: Arthroplasty-fusion is preferred for intervertebral disc degeneration in adjacent upper segments. Fusion-arthroplasty is preferred for patients with lower intervertebral disc degeneration or lower posterior column degeneration. TRIAL REGISTRATION: This research was registered in Chinese Clinical Trial Registry (ChiCTR1900020513).

Mitochondrial dysfunction of induced pluripotent stem cells-based neurodegenerative disease modeling and therapeutic strategy.

Neurodegenerative diseases (NDDs) are disorders in which neurons are lost owing to various factors, resulting in a series of dysfunctions. Their rising prevalence and irreversibility have brought physical pain to patients and economic pressure to both individuals and society. However, the pathogenesis of NDDs has not yet been fully elucidated, hampering the use of precise medication. Induced pluripotent stem cell (IPSC) modeling provides a new method for drug discovery, and exploring the early pathological mechanisms including mitochondrial dysfunction, which is not only an early but a prominent pathological feature of NDDs. In this review, we summarize the iPSC modeling approach of Alzheimer's disease, Parkinson's disease, and Amyotrophic lateral sclerosis, as well as outline typical mitochondrial dysfunction and recapitulate corresponding therapeutic strategies.

Spinal dI4 Interneuron Differentiation From Human Pluripotent Stem Cells.

Spinal interneurons (INs) form intricate local networks in the spinal cord and regulate not only the ascending and descending nerve transduction but also the central pattern generator function. They are therefore potential therapeutic targets in spinal cord injury and diseases. In this study, we devised a reproducible protocol to differentiate human pluripotent stem cells (hPSCs) from enriched spinal dI4 inhibitory GABAergic INs. The protocol is designed based on developmental principles and optimized by using small molecules to maximize its reproducibility. The protocol comprises induction of neuroepithelia, patterning of neuroepithelia to dorsal spinal progenitors, expansion of the progenitors in suspension, and finally differentiation into mature neurons. In particular, we employed both morphogen activators and inhibitors to restrict or "squeeze" the progenitor fate during the stage of neural patterning. We use retinoic acid (RA) which ventralizes cells up to the mid-dorsal region, with cyclopamine (CYC), an SHH inhibitor, to antagonize the ventralization effect of RA, yielding highly enriched dI4 progenitors (90% Ptf1a+, 90.7% Ascl1+). The ability to generate enriched spinal dI4 GABAergicINs will likely facilitate the study of human spinal IN development and regenerative therapies for traumatic injuries and diseases of the spinal cord.

Generation of an induced pluripotent stem cell line HUSTTJi001-A from a Moyamoya disease patient with RNF213 gene mutation.

Moyamoya disease (MMD) is an idiopathic and chronic steno-occlusive cerebrovascular disease. Genetic studies identified RNF213 as a principal susceptibility gene of MMD. In this study, peripheral blood mononuclear cells (PBMCs) were obtained from an MMD patient with RNF213 p. R4810K mutations, and the PBMCs were then reprogrammed to induced pluripotent stem cells (iPSCs) by the transfection of non-integrated episomal vectors. The iPSC line shows pluripotency markers and has the potential for in vitro differentiation into three germ layers, and will be valuable for elucidating the underlying cellular mechanisms of MMD, selecting therapeutic targets, and developing drugs.

MAVS: A Two-Sided CARD Mediating Antiviral Innate Immune Signaling and Regulating Immune Homeostasis.

Mitochondrial antiviral signaling protein (MAVS) functions as a "switch" in the immune signal transduction against most RNA viruses. Upon viral infection, MAVS forms prion-like aggregates by receiving the cytosolic RNA sensor retinoic acid-inducible gene I-activated signaling and further activates/switches on the type I interferon signaling. While under resting state, MAVS is prevented from spontaneously aggregating to switch off the signal transduction and maintain immune homeostasis. Due to the dual role in antiviral signal transduction and immune homeostasis, MAVS has emerged as the central regulation target by both viruses and hosts. Recently, researchers show increasing interest in viral evasion strategies and immune homeostasis regulations targeting MAVS, especially focusing on the post-translational modifications of MAVS, such as ubiquitination and phosphorylation. This review summarizes the regulations of MAVS in antiviral innate immune signaling transduction and immune homeostasis maintenance.

Highly stable graphene oxide composite nanofiltration membrane.

Graphene oxide (GO) based membranes are promising for advanced nanofiltration in water treatments but there is a need to improve water flux and membrane stability. Although the interlayer distance of GO membranes can be expanded using intercalants to improve permeability, achieving uniform intercalation without the added complication of water-induced swelling is challenging. Herein, we report the fabrication of GO hybrid lamellar membranes with controllable layer structures to achieve high performance in nanofiltration. The interlayer spacing of the GO hybrid membrane is regulated using TiO2 intercalants of different sizes, while the stability of GO membranes is enhanced by encapsulating with polyethyleneimine (PEI). The optimal composite membrane delivers a pure water-flux up to 26.0 L m-2 h-1 bar-1 with a 99.9% rejection of methylene blue and eosin under an ultra-low pressure nanofiltration condition. More importantly, the composite membrane sustains good cycling stability after 5 filtration cycles of dye, which enables the potential industrial application in realizing ultra-stable GO based membranes.

Body Part Regression With Self-Supervision.

Body part regression is a promising new technique that enables content navigation through self-supervised learning. Using this technique, the global quantitative spatial location for each axial view slice is obtained from computed tomography (CT). However, it is challenging to define a unified global coordinate system for body CT scans due to the large variabilities in image resolution, contrasts, sequences, and patient anatomy. Therefore, the widely used supervised learning approach cannot be easily deployed. To address these concerns, we propose an annotation-free method named blind-unsupervised-supervision network (BUSN). The contributions of the work are in four folds: (1) 1030 multi-center CT scans are used in developing BUSN without any manual annotation. (2) the proposed BUSN corrects the predictions from unsupervised learning and uses the corrected results as the new supervision; (3) to improve the consistency of predictions, we propose a novel neighbor message passing (NMP) scheme that is integrated with BUSN as a statistical learning based correction; and (4) we introduce a new pre-processing pipeline with inclusion of the BUSN, which is validated on 3D multi-organ segmentation. The proposed method is trained on 1,030 whole body CT scans (230,650 slices) from five datasets, as well as an independent external validation cohort with 100 scans. From the body part regression results, the proposed BUSN achieved significantly higher median R-squared score (=0.9089) than the state-of-the-art unsupervised method (=0.7153). When introducing BUSN as a preprocessing stage in volumetric segmentation, the proposed pre-processing pipeline using BUSN approach increases the total mean Dice score of the 3D abdominal multi-organ segmentation from 0.7991 to 0.8145.

High-resolution 3D abdominal segmentation with random patch network fusion.

Deep learning for three dimensional (3D) abdominal organ segmentation on high-resolution computed tomography (CT) is a challenging topic, in part due to the limited memory provide by graphics processing units (GPU) and large number of parameters and in 3D fully convolutional networks (FCN). Two prevalent strategies, lower resolution with wider field of view and higher resolution with limited field of view, have been explored but have been presented with varying degrees of success. In this paper, we propose a novel patch-based network with random spatial initialization and statistical fusion on overlapping regions of interest (ROIs). We evaluate the proposed approach using three datasets consisting of 260 subjects with varying numbers of manual labels. Compared with the canonical "coarse-to-fine" baseline methods, the proposed method increases the performance on multi-organ segmentation from 0.799 to 0.856 in terms of mean DSC score (p-value < 0.01 with paired t-test). The effect of different numbers of patches is evaluated by increasing the depth of coverage (expected number of patches evaluated per voxel). In addition, our method outperforms other state-of-the-art methods in abdominal organ segmentation. In conclusion, the approach provides a memory-conservative framework to enable 3D segmentation on high-resolution CT. The approach is compatible with many base network structures, without substantially increasing the complexity during inference. Given a CT scan with at high resolution, a low-res section (left panel) is trained with multi-channel segmentation. The low-res part contains down-sampling and normalization in order to preserve the complete spatial information. Interpolation and random patch sampling (mid panel) is employed to collect patches. The high-dimensional probability maps are acquired (right panel) from integration of all patches on field of views.

Phase identification for dynamic CT enhancements with generative adversarial network.

PURPOSE: Dynamic contrast-enhanced computed tomography (CT) is widely used to provide dynamic tissue contrast for diagnostic investigation and vascular identification. However, the phase information of contrast injection is typically recorded manually by technicians, which introduces missing or mislabeling. Hence, imaging-based contrast phase identification is appealing, but challenging, due to large variations among different contrast protocols, vascular dynamics, and metabolism, especially for clinically acquired CT scans. The purpose of this study is to perform imaging-based phase identification for dynamic abdominal CT using a proposed adversarial learning framework across five representative contrast phases. METHODS: A generative adversarial network (GAN) is proposed as a disentangled representation learning model. To explicitly model different contrast phases, a low dimensional common representation and a class specific code are fused in the hidden layer. Then, the low dimensional features are reconstructed following a discriminator and classifier. 36 350 slices of CT scans from 400 subjects are used to evaluate the proposed method with fivefold cross-validation with splits on subjects. Then, 2216 slices images from 20 independent subjects are employed as independent testing data, which are evaluated using multiclass normalized confusion matrix. RESULTS: The proposed network significantly improved correspondence (0.93) over VGG, ResNet50, StarGAN, and 3DSE with accuracy scores 0.59, 0.62, 0.72, and 0.90, respectively (P < 0.001 Stuart-Maxwell test for normalized multiclass confusion matrix). CONCLUSION: We show that adversarial learning for discriminator can be benefit for capturing contrast information among phases. The proposed discriminator from the disentangled network achieves promising results.

Learning from dispersed manual annotations with an optimized data weighting policy.

Purpose: Deep learning methods have become essential tools for quantitative interpretation of medical imaging data, but training these approaches is highly sensitive to biases and class imbalance in the available data. There is an opportunity to increase the available training data by combining across different data sources (e.g., distinct public projects); however, data collected under different scopes tend to have differences in class balance, label availability, and subject demographics. Recent work has shown that importance sampling can be used to guide training selection. To date, these approaches have not considered imbalanced data sources with distinct labeling protocols. Approach: We propose a sampling policy, known as adaptive stochastic policy (ASP), inspired by reinforcement learning to adapt training based on subject, data source, and dynamic use criteria. We apply ASP in the context of multiorgan abdominal computed tomography segmentation. Training was performed with cross validation on 840 subjects from 10 data sources. External validation was performed with 20 subjects from 1 data source. Results: Four alternative strategies were evaluated with the state-of-the-art baseline as upper confident bound (UCB). ASP achieves average Dice of 0.8261 compared to 0.8135 UCB ( p < 0.01 , paired t -test) across fivefold cross validation. On withheld testing datasets, the proposed ASP achieved 0.8265 mean Dice versus 0.8077 UCB ( p < 0.01 , paired t -test). Conclusions: ASP provides a flexible reweighting technique for training deep learning models. We conclude that the proposed method adapts the sample importance, which leverages the performance on a challenging multisite, multiorgan, and multisize segmentation task.

Deep learning assisted differentiation of hepatocellular carcinoma from focal liver lesions: choice of four-phase and three-phase CT imaging protocol.

PURPOSE: To evaluate whether a three-phase dynamic contrast-enhanced CT protocol, when combined with a deep learning model, has similar accuracy in differentiating hepatocellular carcinoma (HCC) from other focal liver lesions (FLLs) compared with a four-phase protocol. METHODS: Three hundred and forty-two patients (mean age 49.1 ± 10.5 years, range 19-86 years, 65.8% male) scanned with a four-phase CT protocol (precontrast, arterial, portal-venous and delayed phases) were retrospectively enrolled. A total of 449 FLLs were categorized into HCC and non-HCC groups based on the best available reference standard. Three convolutional dense networks (CDNs) with the input of four-phase CT images (model A), three-phase images without portal-venous phase (model B) and three-phase images without precontrast phase (model C) were trained on 80% of lesions and evaluated in the other 20% by receiver operating characteristics (ROC) and confusion matrix analysis. The DeLong test was performed to compare the areas under the ROC curves (AUCs) of A with B, B with C, and A with C. RESULTS: The diagnostic accuracy in differentiating HCC from other FLLs on test sets was 83.3% for model A, 81.1% for model B and 85.6% for model C, and the AUCs were 0.925, 0.862 and 0.920, respectively. The AUCs of models A and C did not differ significantly (p = 0.765), but the AUCs of models A and B (p = 0.038) and of models B and C (p = 0.028) did. CONCLUSIONS: When combined with a CDN, a three-phase CT protocol without precontrast showed similar diagnostic accuracy as a four-phase protocol in differentiating HCC from other FLLs, suggesting that the multiphase CT protocol for HCC diagnosis might be optimized by removing the precontrast phase to reduce radiation dose.

The global survival rate among adult out-of-hospital cardiac arrest patients who received cardiopulmonary resuscitation: a systematic review and meta-analysis.

BACKGROUND: To quantitatively summarize the available epidemiological evidence on the survival rate of out-of-hospital cardiac arrest (OHCA) patients who received cardiopulmonary resuscitation (CPR). METHODS: We systematically searched the PubMed, Embase, and Web of Science databases, and the references of retrieved articles were manually reviewed to identify studies reporting the outcome of OHCA patients who received CPR. The overall incidence and outcome of OHCA were assessed using a random-effects meta-analysis. RESULTS: A total of 141 eligible studies were included in this meta-analysis. The pooled incidence of return of spontaneous circulation (ROSC) was 29.7% (95% CI 27.6-31.7%), the rate of survival to hospital admission was 22.0% (95% CI 20.7-23.4%), the rate of survival to hospital discharge was 8.8% (95% CI 8.2-9.4%), the pooled 1-month survival rate was 10.7% (95% CI 9.1-13.3%), and the 1-year survival rate was 7.7% (95% CI 5.8-9.5%). Subgroup analysis showed that survival to hospital discharge was more likely among OHCA patients whose cardiac arrest was witnessed by a bystander or emergency medical services (EMS) (10.5%; 95% CI 9.2-11.7%), who received bystander CPR (11.3%, 95% CI 9.3-13.2%), and who were living in Europe and North America (Europe 11.7%; 95% CI 10.5-13.0%; North America: 7.7%; 95% CI 6.9-8.6%). The survival to discharge (8.6% in 1976-1999 vs. 9.9% in 2010-2019), 1-month survival (8.0% in 2000-2009 vs. 13.3% in 2010-2019), and 1-year survival (8.0% in 2000-2009 vs. 13.3% in 2010-2019) rates of OHCA patients who underwent CPR significantly increased throughout the study period. The Egger's test did not indicate evidence of publication bias for the outcomes of OHCA patients who underwent CPR. CONCLUSIONS: The global survival rate of OHCA patients who received CPR has increased in the past 40 years. A higher survival rate post-OHCA is more likely among patients who receive bystander CPR and who live in Western countries.

Validation and Optimization of Multi-Organ Segmentation on Clinical Imaging Archives.

Segmentation of abdominal computed tomography (CT) provides spatial context, morphological properties, and a framework for tissue-specific radiomics to guide quantitative Radiological assessment. A 2015 MICCAI challenge spurred substantial innovation in multi-organ abdominal CT segmentation with both traditional and deep learning methods. Recent innovations in deep methods have driven performance toward levels for which clinical translation is appealing. However, continued cross-validation on open datasets presents the risk of indirect knowledge contamination and could result in circular reasoning. Moreover, "real world" segmentations can be challenging due to the wide variability of abdomen physiology within patients. Herein, we perform two data retrievals to capture clinically acquired deidentified abdominal CT cohorts with respect to a recently published variation on 3D U-Net (baseline algorithm). First, we retrieved 2004 deidentified studies on 476 patients with diagnosis codes involving spleen abnormalities (cohort A). Second, we retrieved 4313 deidentified studies on 1754 patients without diagnosis codes involving spleen abnormalities (cohort B). We perform prospective evaluation of the existing algorithm on both cohorts, yielding 13% and 8% failure rate, respectively. Then, we identified 51 subjects in cohort A with segmentation failures and manually corrected the liver and gallbladder labels. We re-trained the model adding the manual labels, resulting in performance improvement of 9% and 6% failure rate for the A and B cohorts, respectively. In summary, the performance of the baseline on the prospective cohorts was similar to that on previously published datasets. Moreover, adding data from the first cohort substantively improved performance when evaluated on the second withheld validation cohort.

Semi-Supervised Multi-Organ Segmentation through Quality Assurance Supervision.

Human in-the-loop quality assurance (QA) is typically performed after medical image segmentation to ensure that the systems are performing as intended, as well as identifying and excluding outliers. By performing QA on large-scale, previously unlabeled testing data, categorical QA scores (e.g. "successful" versus "unsuccessful") can be generated. Unfortunately, the precious use of resources for human in-the-loop QA scores are not typically reused in medical image machine learning, especially to train a deep neural network for image segmentation. Herein, we perform a pilot study to investigate if the QA labels can be used as supplementary supervision to augment the training process in a semi-supervised fashion. In this paper, we propose a semi-supervised multi-organ segmentation deep neural network consisting of a traditional segmentation model generator and a QA involved discriminator. An existing 3-D abdominal segmentation network is employed, while the pre-trained ResNet-18 network is used as discriminator. A large-scale dataset of 2027 volumes are used to train the generator, whose 2-D montage images and segmentation mask with QA scores are used to train the discriminator. To generate the QA scores, the 2-D montage images were reviewed manually and coded 0 (success), 1 (errors consistent with published performance), and 2 (gross failure). Then, the ResNet-18 network was trained with 1623 montage images in equal distribution of all three code labels and achieved an accuracy 94% for classification predictions with 404 montage images withheld for the test cohort. To assess the performance of using the QA supervision, the discriminator was used as a loss function in a multi-organ segmentation pipeline. The inclusion of QA-loss function boosted performance on the unlabeled test dataset from 714 patients to 951 patients over the baseline model. Additionally, the number of failures decreased from 606 (29.90%) to 402 (19.83%). The contributions of the proposed method are three-fold: We show that (1) the QA scores can be used as a loss function to perform semi-supervised learning for unlabeled data, (2) the well trained discriminator is learnt by QA score rather than traditional "true/false", and (3) the performance of multi-organ segmentation on unlabeled datasets can be fine-tuned with more robust and higher accuracy than the original baseline method. The use of QA-inspired loss functions represents a promising area of future research and may permit tighter integration of supervised and semi-supervised learning.

Contrast Phase Classification with a Generative Adversarial Network.

Dynamic contrast enhanced computed tomography (CT) is an imaging technique that provides critical information on the relationship of vascular structure and dynamics in the context of underlying anatomy. A key challenge for image processing with contrast enhanced CT is that phase discrepancies are latent in different tissues due to contrast protocols, vascular dynamics, and metabolism variance. Previous studies with deep learning frameworks have been proposed for classifying contrast enhancement with networks inspired by computer vision. Here, we revisit the challenge in the context of whole abdomen contrast enhanced CTs. To capture and compensate for the complex contrast changes, we propose a novel discriminator in the form of a multi-domain disentangled representation learning network. The goal of this network is to learn an intermediate representation that separates contrast enhancement from anatomy and enables classification of images with varying contrast time. Briefly, our unpaired contrast disentangling GAN(CD-GAN) Discriminator follows the ResNet architecture to classify a CT scan from different enhancement phases. To evaluate the approach, we trained the enhancement phase classifier on 21060 slices from two clinical cohorts of 230 subjects. The scans were manually labeled with three independent enhancement phases (non-contrast, portal venous and delayed). Testing was performed on 9100 slices from 30 independent subjects who had been imaged with CT scans from all contrast phases. Performance was quantified in terms of the multi-class normalized confusion matrix. The proposed network significantly improved correspondence over baseline UNet, ResNet50 and StarGAN's performance of accuracy scores 0.54. 0.55, 0.62 and 0.91, respectively (p-value<0.0001 paired t-test for ResNet versus CD-GAN). The proposed discriminator from the disentangled network presents a promising technique that may allow deeper modeling of dynamic imaging against patient specific anatomies.

Outlier Guided Optimization of Abdominal Segmentation.

Abdominal multi-organ segmentation of computed tomography (CT) images has been the subject of extensive research interest. It presents a substantial challenge in medical image processing, as the shape and distribution of abdominal organs can vary greatly among the population and within an individual over time. While continuous integration of novel datasets into the training set provides potential for better segmentation performance, collection of data at scale is not only costly, but also impractical in some contexts. Moreover, it remains unclear what marginal value additional data have to offer. Herein, we propose a single-pass active learning method through human quality assurance (QA). We built on a pre-trained 3D U-Net model for abdominal multi-organ segmentation and augmented the dataset either with outlier data (e.g., exemplars for which the baseline algorithm failed) or inliers (e.g., exemplars for which the baseline algorithm worked). The new models were trained using the augmented datasets with 5-fold cross-validation (for outlier data) and withheld outlier samples (for inlier data). Manual labeling of outliers increased Dice scores with outliers by 0.130, compared to an increase of 0.067 with inliers (p<0.001, two-tailed paired t-test). By adding 5 to 37 inliers or outliers to training, we find that the marginal value of adding outliers is higher than that of adding inliers. In summary, improvement on single-organ performance was obtained without diminishing multi-organ performance or significantly increasing training time. Hence, identification and correction of baseline failure cases present an effective and efficient method of selecting training data to improve algorithm performance.