FedDBL: Communication and Data Efficient Federated Deep-Broad Learning for Histopathological Tissue Classification.

Histopathological tissue classification is a fundamental task in computational pathology. Deep learning (DL)-based models have achieved superior performance but centralized training suffers from the privacy leakage problem. Federated learning (FL) can safeguard privacy by keeping training samples locally, while existing FL-based frameworks require a large number of well-annotated training samples and numerous rounds of communication which hinder their viability in real-world clinical scenarios. In this article, we propose a lightweight and universal FL framework, named federated deep-broad learning (FedDBL), to achieve superior classification performance with limited training samples and only one-round communication. By simply integrating a pretrained DL feature extractor, a fast and lightweight broad learning inference system with a classical federated aggregation approach, FedDBL can dramatically reduce data dependency and improve communication efficiency. Five-fold cross-validation demonstrates that FedDBL greatly outperforms the competitors with only one-round communication and limited training samples, while it even achieves comparable performance with the ones under multiple-round communications. Furthermore, due to the lightweight design and one-round communication, FedDBL reduces the communication burden from 4.6 GB to only 138.4 KB per client using the ResNet-50 backbone at 50-round training. Extensive experiments also show the scalability of FedDBL on model generalization to the unseen dataset, various client numbers, model personalization and other image modalities. Since no data or deep model sharing across different clients, the privacy issue is well-solved and the model security is guaranteed with no model inversion attack risk. Code is available at https://github.com/tianpeng-deng/FedDBL.

Deep learning for colorectal cancer detection in contrast-enhanced CT without bowel preparation: a retrospective, multicentre study.

BACKGROUND: Contrast-enhanced CT scans provide a means to detect unsuspected colorectal cancer. However, colorectal cancers in contrast-enhanced CT without bowel preparation may elude detection by radiologists. We aimed to develop a deep learning (DL) model for accurate detection of colorectal cancer, and evaluate whether it could improve the detection performance of radiologists. METHODS: We developed a DL model using a manually annotated dataset (1196 cancer vs 1034 normal). The DL model was tested using an internal test set (98 vs 115), two external test sets (202 vs 265 in 1, and 252 vs 481 in 2), and a real-world test set (53 vs 1524). We compared the detection performance of the DL model with radiologists, and evaluated its capacity to enhance radiologists' detection performance. FINDINGS: In the four test sets, the DL model had the area under the receiver operating characteristic curves (AUCs) ranging between 0.957 and 0.994. In both the internal test set and external test set 1, the DL model yielded higher accuracy than that of radiologists (97.2% vs 86.0%, p < 0.0001; 94.9% vs 85.3%, p < 0.0001), and significantly improved the accuracy of radiologists (93.4% vs 86.0%, p < 0.0001; 93.6% vs 85.3%, p < 0.0001). In the real-world test set, the DL model delivered sensitivity comparable to that of radiologists who had been informed about clinical indications for most cancer cases (94.3% vs 96.2%, p > 0.99), and it detected 2 cases that had been missed by radiologists. INTERPRETATION: The developed DL model can accurately detect colorectal cancer and improve radiologists' detection performance, showing its potential as an effective computer-aided detection tool. FUNDING: This study was supported by National Science Fund for Distinguished Young Scholars of China (No. 81925023); Regional Innovation and Development Joint Fund of National Natural Science Foundation of China (No. U22A20345); National Natural Science Foundation of China (No. 82072090 and No. 82371954); Guangdong Provincial Key Laboratory of Artificial Intelligence in Medical Image Analysis and Application (No. 2022B1212010011); High-level Hospital Construction Project (No. DFJHBF202105).

Revisiting the steatosis-associated fibrosis estimator score in young Asian subjects with steatotic liver disease and consideration for population variability.

BACKGROUND AND AIM: The steatosis-associated fibrosis estimator (SAFE) score has been developed to distinguish clinically significant fibrosis in patients with steatotic liver disease (SLD). However, validation of its performance in Asian subjects is limited. This study aimed to evaluate the performance of the SAFE score in Asian subjects with biopsy-proven SLD and in different subgroups according to age, sex, and body mass index. METHODS: We retrospectively analyzed 6383 living liver donors who underwent a liver biopsy between 2005 and 2023. Of these, 1551 subjects with biopsy-proven SLD were included. The performance of the SAFE score was evaluated using areas under the curve and compared with those of the nonalcoholic fatty liver disease fibrosis score (NFS) and fibrosis-4 index (FIB-4). RESULTS: The prevalence of clinically significant fibrosis in the cohort was 2.2%. The proportion of subjects with a "low-risk" SAFE score was the highest (91.0%), followed by those with "intermediate-risk" (7.8%) and "high-risk" (1.2%) scores. The prevalence of fibrosis in subjects with low-risk, intermediate-risk, and high-risk scores was 1.6%, 6.6%, and 21.1%, respectively. The SAFE outperformed FIB-4 and NFS (area under the curve: 0.70 vs 0.64 for both NFS and FIB-4). However, it showed low diagnostic accuracy and sensitivity (27%) at the low cutoff (SAFE < 0) in subjects aged 30-39 years (fibrosis: 1.2%), despite having a high negative predictive value (0.99). CONCLUSION: While the SAFE score demonstrates superior performance compared with other noninvasive tests in Asian subjects with SLD, its performance varies across age groups. In younger subjects, particularly, its performance may be more limited.

Updated Reference Intervals for Alanine Aminotransferase in a Metabolically and Histologically Normal Population.

BACKGROUND & AIMS: This study aims to reevaluate upper reference limit (URL) for alanine aminotransferase (ALT) by considering the changing epidemiology of major liver diseases. We employed histological and metabolic parameters in Asian living liver donors. METHODS: We performed a retrospective analysis of 5455 potential living liver donors from 2005 to 2019. Participants were screened for hepatitis B, C, HIV, and alcohol use. Histologically and metabolically healthy participants were assessed using the Prati criteria (body mass index <23 kg/m2, triglyceride ≤200 mg/dL, fasting glucose ≤105 mg/dL, total cholesterol ≤220 mg/dL). The updated ALT-URL was determined as the 95th percentile among participants without hepatic steatosis and who met the Prati criteria. RESULTS: The median age was 30 years, with a male predominance (66.2%). Among 5455 participants, 3162 (58.0%) showed no hepatic steatosis, with 1553 (49.1%) meeting both the criteria for no steatosis and the Prati criteria for metabolic health. The updated URL for ALT in these participants was 34 U/L for males and 22 U/L for females, which was significantly lower than conventionally accepted values. Using this revised ALT-URL, 72.8% of males with ALT levels ≥34 U/L and 55.0% of females with ALT levels ≥22 U/L showed signs of steatosis, whereas 32.7% of males and 22.2% of females met the criteria for metabolic syndrome. CONCLUSIONS: Our study provided the newly established reference intervals for ALT levels in a metabolically and histologically verified Asian population. The proposed URL for ALT are 34 U/L and 22 U/L for males and females, respectively.

FedDUS: Lung tumor segmentation on CT images through federated semi-supervised with dynamic update strategy.

BACKGROUND AND OBJECTIVE: Lung tumor annotation is a key upstream task for further diagnosis and prognosis. Although deep learning techniques have promoted automation of lung tumor segmentation, there remain challenges impeding its application in clinical practice, such as a lack of prior annotation for model training and data-sharing among centers. METHODS: In this paper, we use data from six centers to design a novel federated semi-supervised learning (FSSL) framework with dynamic model aggregation and improve segmentation performance for lung tumors. To be specific, we propose a dynamically updated algorithm to deal with model parameter aggregation in FSSL, which takes advantage of both the quality and quantity of client data. Moreover, to increase the accessibility of data in the federated learning (FL) network, we explore the FAIR data principle while the previous federated methods never involve. RESULT: The experimental results show that the segmentation performance of our model in six centers is 0.9348, 0.8436, 0.8328, 0.7776, 0.8870 and 0.8460 respectively, which is superior to traditional deep learning methods and recent federated semi-supervised learning methods. CONCLUSION: The experimental results demonstrate that our method is superior to the existing FSSL methods. In addition, our proposed dynamic update strategy effectively utilizes the quality and quantity information of client data and shows efficiency in lung tumor segmentation. The source code is released on (https://github.com/GDPHMediaLab/FedDUS).

A Colorectal Coordinate-Driven Method for Colorectum and Colorectal Cancer Segmentation in Conventional CT Scans.

Automated colorectal cancer (CRC) segmentation in medical imaging is the key to achieving automation of CRC detection, staging, and treatment response monitoring. Compared with magnetic resonance imaging (MRI) and computed tomography colonography (CTC), conventional computed tomography (CT) has enormous potential because of its broad implementation, superiority for the hollow viscera (colon), and convenience without needing bowel preparation. However, the segmentation of CRC in conventional CT is more challenging due to the difficulties presenting with the unprepared bowel, such as distinguishing the colorectum from other structures with similar appearance and distinguishing the CRC from the contents of the colorectum. To tackle these challenges, we introduce DeepCRC-SL, the first automated segmentation algorithm for CRC and colorectum in conventional contrast-enhanced CT scans. We propose a topology-aware deep learning-based approach, which builds a novel 1-D colorectal coordinate system and encodes each voxel of the colorectum with a relative position along the coordinate system. We then induce an auxiliary regression task to predict the colorectal coordinate value of each voxel, aiming to integrate global topology into the segmentation network and thus improve the colorectum's continuity. Self-attention layers are utilized to capture global contexts for the coordinate regression task and enhance the ability to differentiate CRC and colorectum tissues. Moreover, a coordinate-driven self-learning (SL) strategy is introduced to leverage a large amount of unlabeled data to improve segmentation performance. We validate the proposed approach on a dataset including 227 labeled and 585 unlabeled CRC cases by fivefold cross-validation. Experimental results demonstrate that our method outperforms some recent related segmentation methods and achieves the segmentation accuracy in DSC for CRC of 0.669 and colorectum of 0.892, reaching to the performance (at 0.639 and 0.890, respectively) of a medical resident with two years of specialized CRC imaging fellowship.

Outcomes of Liver Resection and Transarterial Chemoembolization in Patients with Multinodular BCLC-A Hepatocellular Carcinoma.

Background: This study aimed to compare the outcomes of liver resection (LR) and transarterial chemoembolization (TACE) in patients with multinodular hepatocellular carcinoma (HCC) within the Milan criteria who were not eligible for liver transplantation. Methods: We retrospectively analyzed 483 patients with multinodular HCC within the Milan criteria, who underwent either LR or TACE as an initial therapy between 2013 and 2022. The overall survival (OS) in the entire population and recurrence-free survival (RFS) in patients who underwent LR and TACE and achieved a complete response were analyzed. Propensity score (PS) matching analysis was also used for a fair comparison of outcomes between the two groups. Results: Among the 483 patients, 107 (22.2%) and 376 (77.8%) underwent LR and TACE, respectively. The median size of the largest tumor was 2.0 cm, and 72.3% of the patients had two HCC lesions. The median OS and RFS were significantly longer in the LR group than in the TACE group (p <0.01 for both). In the multivariate analysis, TACE (adjusted hazard ratio [aHR], 1.81 and aHR, 2.41) and large tumor size (aHR, 1.43 and aHR, 1.44) were significantly associated with worse OS and RFS, respectively. The PS-matched analysis also demonstrated that the LR group had significantly longer OS and RFS than the TACE group (PS <0.05). Conclusion: In this study, LR showed better OS and RFS than TACE in patients with multinodular Barcelona Clinic Liver Cancer stage A HCC. Therefore, LR can be considered an effective treatment option for these patients.

The associations between fibrosis changes and liver-related events in patients with metabolic dysfunction-associated steatotic liver disease.

BACKGROUND: The prognosis of metabolic dysfunction-associated steatotic liver disease (MASLD) is associated with liver fibrosis. We investigated the associations between changes in liver stiffness measurement (LSM) over 3-year period and the development of cirrhosis or hepatocellular carcinoma (HCC) in patients with MASLD. METHODS: This study involved patients with MASLD who underwent transient elastography at baseline and 3 years after baseline from 2012 to 2020. Low (L), indeterminate (I) and high (H) LSM values were classified as <8 kPa, 8-12 kPa and >12 kPa respectively. RESULTS: Among 1738 patients, 150 (8.6%) were diagnosed with cirrhosis or HCC. The proportions of patients with L, I and H risk were 69.7%, 19.9% and 10.5% at baseline, and 78.8%, 12.8% and 8.4% at 3 years after baseline, respectively. The incidence rates of cirrhosis or HCC per 1000 person-years were 3.7 (95% confidence interval [CI], 2.4-5.5) in the L → L + I group, 23.9 (95% CI, 17.1-32.6) in the I → L + I group, 38.3 (95% CI, 22.3-61.3) in the H → L + I group, 62.5 (95% CI, 32.3-109.2) in the I → H group, 67.8 (95% CI, 18.5-173.6) in the L → H group and 93.9 (95% CI 70.1-123.1) in the H → H group. Two risk factors for the development of cirrhosis or HCC were LSM changes and low platelet counts. CONCLUSION: LSM changes could predict clinical outcomes in patients with MASLD. Thus, it is important to monitor changes in the fibrotic burden by regular assessment of LSM values.

Development and Validation of a Deep Learning Model to Reduce the Interference of Rectal Artifacts in MRI-based Prostate Cancer Diagnosis.

Purpose To develop an MRI-based model for clinically significant prostate cancer (csPCa) diagnosis that can resist rectal artifact interference. Materials and Methods This retrospective study included 2203 male patients with prostate lesions who underwent biparametric MRI and biopsy between January 2019 and June 2023. Targeted adversarial training with proprietary adversarial samples (TPAS) strategy was proposed to enhance model resistance against rectal artifacts. The automated csPCa diagnostic models trained with and without TPAS were compared using multicenter validation datasets. The impact of rectal artifacts on the diagnostic performance of each model at the patient and lesion levels was compared using the area under the receiver operating characteristic curve (AUC) and the area under the precision-recall curve (AUPRC). The AUC between models was compared using the DeLong test, and the AUPRC was compared using the bootstrap method. Results The TPAS model exhibited diagnostic performance improvements of 6% at the patient level (AUC: 0.87 vs 0.81, P < .001) and 7% at the lesion level (AUPRC: 0.84 vs 0.77, P = .007) compared with the control model. The TPAS model demonstrated less performance decline in the presence of rectal artifact-pattern adversarial noise than the control model (ΔAUC: -17% vs -19%, ΔAUPRC: -18% vs -21%). The TPAS model performed better than the control model in patients with moderate (AUC: 0.79 vs 0.73, AUPRC: 0.68 vs 0.61) and severe (AUC: 0.75 vs 0.57, AUPRC: 0.69 vs 0.59) artifacts. Conclusion This study demonstrates that the TPAS model can reduce rectal artifact interference in MRI-based csPCa diagnosis, thereby improving its performance in clinical applications. Keywords: MR-Diffusion-weighted Imaging, Urinary, Prostate, Comparative Studies, Diagnosis, Transfer Learning Clinical trial registration no. ChiCTR23000069832 Supplemental material is available for this article. Published under a CC BY 4.0 license.

A computed tomography-based multitask deep learning model for predicting tumour stroma ratio and treatment outcomes in patients with colorectal cancer: a multicentre cohort study.

BACKGROUND: Tumour-stroma interactions, as indicated by tumour-stroma ratio (TSR), offer valuable prognostic stratification information. Current histological assessment of TSR is limited by tissue accessibility and spatial heterogeneity. The authors aimed to develop a multitask deep learning (MDL) model to noninvasively predict TSR and prognosis in colorectal cancer (CRC). MATERIALS AND METHODS: In this retrospective study including 2268 patients with resected CRC recruited from four centres, the authors developed an MDL model using preoperative computed tomography (CT) images for the simultaneous prediction of TSR and overall survival. Patients in the training cohort ( n =956) and internal validation cohort (IVC, n =240) were randomly selected from centre I. Patients in the external validation cohort 1 (EVC1, n =509), EVC2 ( n =203), and EVC3 ( n =360) were recruited from other three centres. Model performance was evaluated with respect to discrimination and calibration. Furthermore, the authors evaluated whether the model could predict the benefit from adjuvant chemotherapy. RESULTS: The MDL model demonstrated strong TSR discrimination, yielding areas under the receiver operating curves (AUCs) of 0.855 (95% CI, 0.800-0.910), 0.838 (95% CI, 0.802-0.874), and 0.857 (95% CI, 0.804-0.909) in the three validation cohorts, respectively. The MDL model was also able to predict overall survival and disease-free survival across all cohorts. In multivariable Cox analysis, the MDL score (MDLS) remained an independent prognostic factor after adjusting for clinicopathological variables (all P <0.05). For stage II and stage III disease, patients with a high MDLS benefited from adjuvant chemotherapy [hazard ratio (HR) 0.391 (95% CI, 0.230-0.666), P =0.0003; HR=0.467 (95% CI, 0.331-0.659), P <0.0001, respectively], whereas those with a low MDLS did not. CONCLUSION: The multitask DL model based on preoperative CT images effectively predicted TSR status and survival in CRC patients, offering valuable guidance for personalized treatment. Prospective studies are needed to confirm its potential to select patients who might benefit from chemotherapy.

Improving the hepatitis C virus care cascade with the in-hospital Reflex tEsting ALarm-C (REAL-C) model.

BACKGROUND: The World Health Organization (WHO) has set targets to eliminate viral hepatitis, including hepatitis C virus (HCV) infection, by 2030. We present the results of the in-hospital Reflex tEsting ALarm-C (REAL-C) model, which incorporates reflex HCV RNA testing and sending alerts to physicians. METHODS: We conducted a retrospective study analysing the data of 1730 patients who newly tested positive for anti-HCV between March 2020 and June 2023. Three distinct periods were defined: pre-REAL-C (n = 696), incomplete REAL-C (n = 515) and complete REAL-C model periods (n = 519). The primary outcome measure was the HCV RNA testing rate throughout the study period. Additionally, we assessed the referral rate to the gastroenterology department, linkage time for diagnosis and treatment and the treatment rate. RESULTS: The rate of HCV RNA testing increased significantly from 51.0% (pre-REAL-C) to 95.6% (complete REAL-C). This improvement was consistent across clinical departments, regardless of patients' comorbidities. Among patients with confirmed HCV infection, the gastroenterology referral rate increased from 57.1% to 81.1% after the REAL-C model. The treatment rate among treatment-eligible patients was 92.4% during the study period. The mean interval from anti-HCV positivity to HCV RNA testing decreased from 45.1 to 1.9 days. The mean interval from the detection of anti-HCV positivity to direct-acting antiviral treatment also decreased from 89.5 to 49.5 days with the REAL-C model. CONCLUSION: The REAL-C model, featuring reflex testing and physician alerts, effectively increased HCV RNA testing rates and streamlined care cascades. Our model facilitated progress towards achieving WHO's elimination goals for HCV infection.

Regorafenib plus nivolumab in unresectable hepatocellular carcinoma: the phase 2 RENOBATE trial.

Regorafenib has anti-tumor activity in patients with unresectable hepatocellular carcinoma (uHCC) with potential immunomodulatory effects, suggesting that its combination with immune checkpoint inhibitor may have clinically meaningful benefits in patients with uHCC. The multicenter, single-arm, phase 2 RENOBATE trial tested regorafenib-nivolumab as front-line treatment for uHCC. Forty-two patients received nivolumab 480 mg every 4 weeks and regorafenib 80 mg daily (3-weeks-on/1-week-off schedule). The primary endpoint was the investigator-assessed objective response rate (ORR) per Response Evaluation Criteria in Solid Tumors (RECIST) version 1.1. The secondary endpoints included safety, progression-free survival (PFS) and overall survival (OS). ORR per RECIST version 1.1 was 31.0%, meeting the primary endpoint. The most common adverse events were palmar-plantar erythrodysesthesia syndrome (38.1%), alopecia (26.2%) and skin rash (23.8%). Median PFS was 7.38 months. The 1-year OS rate was 80.5%, and the median OS was not reached. Exploratory single-cell RNA sequencing analyses of peripheral blood mononuclear cells showed that long-term responders exhibited T cell receptor repertoire diversification, enrichment of genes representing immunotherapy responsiveness in MKI67+ proliferating CD8+ T cells and a higher probability of M1-directed monocyte polarization. Our data support further clinical development of the regorafenib-nivolumab combination as front-line treatment for uHCC and provide preliminary insights on immune biomarkers of response. ClinicalTrials.gov identifier: NCT04310709 .

High-strength and crack-free welding of 2024 aluminium alloy via Zr-core-Al-shell wire.

The 2000 series aluminium alloys are qualified for widespread use in lightweight structures, but solidification cracking during fusion welding has been a long-standing issue. Here, we create a zirconium (Zr)-core-aluminium (Al)-shell wire (ZCASW) and employ the oscillating laser-arc hybrid welding technique to control solidification during welding, and ultimately achieve reliable and crack-free welding of 2024 aluminium alloy. We select Zr wires with an ideal lattice match to Al based on crystallographic information and wind them by the Al wires with similar chemical components to the parent material. Crack-free, equiaxed (where the length, width and height of the grains are roughly equal), fine-grained microstructures are acquired, thereby considerably increasing the tensile strength over that of conventional fusion welding joints, and even comparable to that of friction stir welding joints. This work has important engineering application value in welding of high-strength aluminum alloys.

SwinHR: Hemodynamic-powered hierarchical vision transformer for breast tumor segmentation.

Accurate and automated segmentation of breast tumors in dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) plays a critical role in computer-aided diagnosis and treatment of breast cancer. However, this task is challenging, due to random variation in tumor sizes, shapes, appearances, and blurred boundaries of tumors caused by inherent heterogeneity of breast cancer. Moreover, the presence of ill-posed artifacts in DCE-MRI further complicate the process of tumor region annotation. To address the challenges above, we propose a scheme (named SwinHR) integrating prior DCE-MRI knowledge and temporal-spatial information of breast tumors. The prior DCE-MRI knowledge refers to hemodynamic information extracted from multiple DCE-MRI phases, which can provide pharmacokinetics information to describe metabolic changes of the tumor cells over the scanning time. The Swin Transformer with hierarchical re-parameterization large kernel architecture (H-RLK) can capture long-range dependencies within DCE-MRI while maintaining computational efficiency by a shifted window-based self-attention mechanism. The use of H-RLK can extract high-level features with a wider receptive field, which can make the model capture contextual information at different levels of abstraction. Extensive experiments are conducted in large-scale datasets to validate the effectiveness of our proposed SwinHR scheme, demonstrating its superiority over recent state-of-the-art segmentation methods. Also, a subgroup analysis split by MRI scanners, field strength, and tumor size is conducted to verify its generalization. The source code is released on (https://github.com/GDPHMediaLab/SwinHR).

Response to atezolizumab plus bevacizumab specific for lung and lymph node metastases affects survival of patients with HCC.

BACKGROUND & AIMS: Tumour microenvironment heterogeneity among different organs can influence immunotherapy responses. Here, we evaluated the impact of differential organ-specific responses on survival in patients with advanced-stage hepatocellular carcinoma (HCC) treated with atezolizumab plus bevacizumab (Atezo/Bev). METHODS: We retrospectively analysed 366 consecutive patients with advanced-stage HCC treated with Atezo/Bev as first-line systemic treatment. Therapeutic response was assessed using RECIST v1.1. Patients were divided into an intention-to-treat (ITT) group (patients treated with ≥1 dose of Atezo/Bev) and a per-protocol (PP) analysis group (patients with at least one measurable lesion irrespective of location treated with ≥3 doses of Atezo/Bev). Overall response and organ-specific response at initial and best response were evaluated in the PP group. Responders were defined as patients achieving complete remission or partial response. Initial progressors were defined as patients with progressive disease after three doses of Atezo/Bev. RESULTS: The ITT and PP groups comprised 324 and 236 patients, respectively. In the PP group, the organ-specific response rate of lung and lymph node (LN) metastases at both initial and best responses were higher than those of intrahepatic lesions and macrovascular tumour thrombosis. Lung and LN-specific response rates were 21.1% and 23.5%, respectively, at initial response, and 24.7% and 31.4%, respectively, at best response. Both initial pulmonary and lymphatic progressors (adjusted hazard ratios [95% confidence intervals], 6.37 [2.10-19.3], and 8.36 [2.16-32.4], respectively) were independently associated with survival regardless of intrahepatic response. CONCLUSIONS: The response of metastatic HCC to the Atezo/Bev regimen may be used to determine whether to continue treatment or switch to second-line treatment at an early phase of therapy.

Vision Transformers for Computational Histopathology.

Computational histopathology is focused on the automatic analysis of rich phenotypic information contained in gigabyte whole slide images, aiming at providing cancer patients with more accurate diagnosis, prognosis, and treatment recommendations. Nowadays deep learning is the mainstream methodological choice in computational histopathology. Transformer, as the latest technological advance in deep learning, learns feature representations and global dependencies based on self-attention mechanisms, which is increasingly gaining prevalence in this field. This article presents a comprehensive review of state-of-the-art vision transformers that have been explored in histopathological image analysis for classification, segmentation, and survival risk regression applications. We first overview preliminary concepts and components built into vision transformers. Various recent applications including whole slide image classification, histological tissue component segmentation, and survival outcome prediction with tailored transformer architectures are then discussed. We finally discuss key challenges revolving around the use of vision transformers and envisioned future perspectives. We hope that this review could provide an elaborate guideline for readers to explore vision transformers in computational histopathology, such that more advanced techniques assisting in the precise diagnosis and treatment of cancer patients could be developed.

Prediction of Hepatocellular Carcinoma Development in Korean Patients after Hepatitis C Cure with Direct-Acting Antivirals.

Background/Aims: With the wide application of direct-acting antivirals (DAAs) for hepatitis C virus infection, the number of patients achieving a sustained virologic response (SVR) will continue to increase. However, no consensus has been achieved on exempting SVR-achieving patients from hepatocellular carcinoma (HCC) surveillance. Methods: Between 2013 and 2021, 873 Korean patients who achieved SVR following DAA treatment were analyzed. We evaluated the predictive performance of seven noninvasive scores (PAGE-B, modified PAGE-B, Toronto HCC risk index, fibrosis-4, aspartate aminotransferase-to-platelet ratio index, albumin-bilirubin, and age male albumin-bilirubin platelet [aMAP]) at baseline and after SVR. Results: The mean age of the 873 patients (39.3% males) was 59.1 years, and 224 patients (25.7%) had cirrhosis. During 3,542 person-years of follow-up, 44 patients developed HCC, with an annual incidence of 1.24/100 person-years. Male sex (adjusted hazard ratio [AHR], 2.21), cirrhosis (AHR, 7.93), and older age (AHR, 1.05) were associated with a significantly higher HCC risk in multivariate analysis. The performance of all scores at the time of SVR were numerically better than those at baseline as determined by the integrated area under the curve. Time-dependent area under the curves for predicting the 3-, 5-, and 7-year risk of HCC after SVR were higher in mPAGE-B (0.778, 0.746, and 0.812, respectively) and aMAP (0.776, 0.747, and 0.790, respectively) systems than others. No patients predicted as low-risk by the aMAP or mPAGE-B systems developed HCC. Conclusions: aMAP and mPAGE-B scores demonstrated the highest predictive performance for de novo HCC in DAA-treated, SVR-achieving patients. Hence, these two systems may be used to identify low-risk patients that can be exempted from HCC surveillance.

Longitudinal changes in renal function in patients with chronic hepatitis B on antiviral treatment.

BACKGROUND: Patients with chronic hepatitis B (CHB) on nucleos(t)ide analogues (NUCs) often experience renal function decline. Conflicting results regarding the impact of NUC use and renal function have recently been reported. AIM: To examine longitudinal changes in renal function according to the NUC treatment type compared with untreated patients METHODS: From 2014 to 2022, we retrospectively analysed 10,642 patients with CHB. The primary outcome was chronic kidney disease (CKD) progression, which was defined as a minimum one-stage elevation. We applied propensity score (PS) matching for outcome comparisons. RESULTS: In the PS-matched cohort of 1996 pairs, the NUC-treated group (7.6/100 person-years [PYs]) had a significantly higher CKD progression risk than the untreated group (4.4/100 PYs), with a hazard ratio (HR) of 1.70 (p < 0.001). The tenofovir disoproxil fumarate (TDF)-treated group (7.9/100 PYs) showed a 1.76-fold increased CKD progression risk compared with the untreated group (4.5/100 PYs) in the PS-matched cohort (p < 0.001). Both the entecavir- and tenofovir alafenamide (TAF)-treated groups showed CKD progression risks comparable to those of the untreated group in the PS-matched cohorts of 755 and 426 pairs, respectively (p = 0.132 and p = 0.120, respectively). No significant CKD progression risk was found between the entecavir- (6.0/100 PYs) and TAF-treated (5.2/100 PYs) groups in the PS-matched cohort of 510 pairs (p = 0.118). CONCLUSIONS: NUC-treated patients, especially those on TDF, faced a higher CKD progression risk than untreated patients. Entecavir- and TAF-treated patients had comparable CKD progression risks to untreated patients. No difference was observed between entecavir and TAF in the risk of CKD progression.

Cardiovascular risk in chronic hepatitis B patients treated with tenofovir disoproxil fumarate or tenofovir alafenamide.

BACKGROUND/AIMS: Tenofovir disoproxil fumarate (TDF) is known to have a lipid-lowering effect. This is in contrast to tenofovir alafenamide (TAF), which has a lipid-neutral effect. Therefore, concerns have been raised as to whether these differences affect long-term cardiovascular risk. Here, we aimed to evaluate the long-term risk of cardiovascular events in chronic hepatitis B (CHB) patients treated with TAF or TDF. METHODS: We retrospectively analyzed 4,124 treatment-naïve CHB patients treated with TDF (n=3,186) or TAF (n=938) between 2012 and 2022. The primary outcome was a composite endpoint of major adverse cardiovascular events (MACE), including myocardial infarction, ischemic stroke, and hospitalization for unstable angina or heart failure. Serial changes in lipid profiles between two treatments were also explored. RESULTS: The median age of the patients was 50.6 years, and 60.6% of the patients were male. At baseline, 486 (11.8%) and 637 (15.4%) of the patients had dyslipidemia and fatty liver, respectively. A total of 42 MACE occurred, with an annual incidence of 0.2%/100 person-years (PYs). At 1, 3, and 5 years, the cumulative risk of MACE was 0.4%, 0.8%, and 1.2% in patients treated with TDF, and 0.2%, 0.7%, and 0.7% in patients treated with TAF, respectively (p=0.538). No significant differences in the risk of MACE were observed between TDF and TAF. A multivariable analysis found that current smoker and a history of cardiovascular events were risk factors associated with an increased risk of MACE. CONCLUSION: Patients treated with TAF had comparable risks of cardiovascular outcomes, defined as MACE, as patients treated with TDF.

A risk prediction model for hepatocellular carcinoma in non-alcoholic fatty liver disease without cirrhosis: HCC prediction for non-cirrhotic NAFLD.

BACKGROUND & AIMS: Although non-alcoholic fatty liver disease (NAFLD) is becoming a leading cause of hepatocellular carcinoma (HCC), HCC risk in non-cirrhotic NAFLD received little attention. We aimed to develop and validate an HCC risk prediction model for non-cirrhotic NAFLD. METHODS: A nationwide cohort of non-cirrhotic NAFLD patients in Korea was recruited to develop a risk prediction model and validate it internally (n = 409 088). A model using a simplified point system was developed by Cox proportional hazard model. K-fold cross-validation assessed the accuracy, discrimination and calibration. The model was validated externally using a hospital cohort from Asan Medical Center (n = 8721). RESULTS: An 11-point HCC risk prediction model for non-cirrhotic NAFLD was developed using six independent factors of age, sex, diabetes, obesity, serum alanine aminotransferase level and gamma-glutamyl transferase level (c-index 0.75). The average area under receiver operating curves (AUROCs) of the model was 0.72 at 5 years and 0.75 at 10 years. In the external validation cohort, the AUROCs were 0.79 [95% confidence interval [CI], 0.59-0.95] at 5 years and 0.84 (95% CI, 0.73-0.94) at 10 years. The calibration plots showed the expected risks corresponded well with the observed risks. Risk stratification categorized patients into the low (score 0-6), moderate (7, 8) and high (9-11; estimated incidence rate >0.2%/year) risk groups. CONCLUSIONS: A novel HCC risk prediction model for non-cirrhotic NAFLD patients was developed and validated with fair performance. The model is expected to serve as a simple and reliable tool to assess HCC risk and assist precision screening of HCC.