Dennis E. Anderson, Mario Keko, Joanna James
et al.
ABSTRACT Purpose This study investigated the effect of bone metastasis on the biomechanical environment of human vertebrae in patients with metastatic spine disease through the metric of load‐to‐strength ratio (LSR). Specifically, we compared the patients' LSRs to age and sex‐similar noncancer controls from the Framingham Heart Study. Methods Derived from clinical CT data of 135 metastatic spine disease patients planned for radiotherapy and 246 normative controls from the Framingham Heart Study, individualized spinal musculoskeletal models and vertebral strength estimates were used to compute level‐specific LSR under natural standing and three weight‐holding conditions (standing + weight, flexion + weight, and lateral bending + weight). Results Adjusted for age, BMI, and spinal region, osteosclerotic and mixed lesion vertebrae had higher strength than osteolytic and control vertebrae. The musculoskeletal models suggested breast, prostate, and male lung cancer patients had higher compressive vertebral loading, and female lung cancer patients had lower compressive vertebral loading than controls. Male patients had higher standardized LSRs in natural standing, while female patients had lower LSRs for all activities than controls. Independent of sex, vertebrae with osteosclerotic and mixed bone metastasis had lower LSRs than controls, while, for osteolytic bone lesions, males had higher and females lower LSRs than controls. Vertebrae with no observed lesion on CT had higher LSRs than controls in males and lower LSRs in females. Discussion Our findings highlighted that primary cancer and lesion type differentially affected task‐specific vertebral loading and strength, thus modifying the vertebral LSRs. Sex‐mediated differences in LSRs between FHS controls and vertebrae with no observed metastatic lesions suggest that considering the latter as “normal” should be taken with care. Our initial assessment supports further examination of whether vertebral LSR measurements are associated with vertebral risk and, if so, what threshold values indicate risk. Level of Evidence 3.
Lewis L. Shi, MD, James S. MacLeod, MD, Nicholas H. Maassen, MD
et al.
Hypothesis/Background: This study compares our “angle” method to the ''best-fit circle'' method and examines existing literature on glenoid bone loss measurement. We hypothesized that the “angle” method of calculating glenoid bone loss would perform comparably to the “best-fit circle” method. Methods: The “angle” method calculates bone loss by subtracting the area of the triangle placed along the length of the defect from the sector subtended by the arc of the bone defect and is an adaptation of the Lederman ratio to estimate glenoid bone loss. To validate our method, we measured glenoid bone loss on 26 shoulder computed tomography scans using both the “angle” and “best-fit circle” methods. All patients had the diagnosis of anterior shoulder instability and had a computed tomography arthrogram performed due to suspected bone loss; those with glenohumeral arthritis were excluded. Results: Twenty-six patients were included. The glenoid bone loss measured using the “angle” method was 17.5% (standard deviation [SD], 7) on average, while the “best-fit circle” method yielded 23.2% (SD, 7). The mean difference between the 2 methods was 5.4% (P ≤ .0001; SD, 2.1), with the “best-fit circle” method finding higher average bone loss. Intra-rater reliability was excellent for the “angle” method, with a correlation coefficient of 0.965-0.995 compared to 0.953 and 0.992 for the “best-fit circle” method. Conclusion: The “angle” method of calculating glenoid bone loss predicts a 5.4% decreased absolute bone loss than the commonly used “best-fit circle” method and shows high intra-rater reliability, with a correlation coefficient of 0.965-0.995.
Orthopedic surgery, Diseases of the musculoskeletal system
Priyanka Gupta, Mohan Babu Nema, Arvind Karoria
et al.
Background:
Effective post-operative pain management is essential for early recovery and patient satisfaction following knee arthroscopy. This study aimed to evaluate the post-operative analgesic efficacy and safety of ultrasound-guided adductor canal block (ACB) compared to conventional intravenous morphine analgesia.
Materials and Methods:
This randomized, controlled, interventional study was conducted in the post-anesthesia care unit (PACU) of the Department of Anesthesiology in an Indian Hospital. Eighty adult patients (American Society of Anesthesiologists [ASA] I–II) undergoing unilateral knee arthroscopy under general anesthesia were randomly divided into two groups: Group M received intravenous morphine (0.1 mg/kg) before incision, and Group B received an ultrasound-guided ACB with 15 mL of 0.25% bupivacaine before extubation. Post-operative analgesic efficacy was assessed by the requirement of rescue analgesia and the time to achieve a Visual Analog Score (VAS) <3. Adverse effects and antiemetic requirements were also recorded. Statistical analysis was performed using the Statistical Package for Social Sciences version 17.0, and a P < 0.05 was considered significant.
Results:
Both groups were comparable in terms of age, sex, ASA physical status, and pre-operative vitals (P > 0.05). Rescue analgesia in the PACU was required in 47.5% of patients in Group M and 10.0% in Group B (P < 0.001). The mean time to achieve VAS <3 was significantly shorter in Group B (11.00 ± 3.79 min) compared to Group M (16.00 ± 9.00 min) (P = 0.002). The requirement of antiemetic medication was lower in Group B (20.0%) than in Group M (42.5%) (P = 0.030). No adverse events were reported in either group.
Conclusion:
Ultrasound-guided ACB provides superior post-operative analgesia, faster pain relief, and fewer side effects compared to intravenous morphine in patients undergoing knee arthroscopy.
Orthopedic surgery, Diseases of the musculoskeletal system
Alexandra Prado-Mantilla, Joseph Sheheen, Julie Underwood
et al.
Loss-of-function studies are a central approach to understanding gene/protein function. In mice, this often relies upon heritable recombination at the DNA level. This approach is slow and nonreversible, which limits both spatial and temporal resolution of analysis. Recently, degron techniques that directly target proteins for degradation have been successfully used to quickly and reversibly knock down proteins. Currently, these systems have been limited by lack of tissue/cell type specificity. Here, we generated mice that allow spatial and temporal control of GFP-tagged protein degradation. This DegronGFP line leads to degradation of GFP-tagged proteins in different cellular compartments and in distinct cell types. Further, it is rapid and reversible. We used DegronGFP to probe the function of the glucocorticoid receptor in the epidermis and demonstrate that it has distinct functions in proliferative and differentiated cells—an analysis that would not have been possible with traditional recombination approaches. We propose that the ability to use GFP knock-in lines for loss-of-function analysis will provide additional motivation for generation of these useful tools.
Abstract Osteoclasts (OCs) differentiate from macrophages in response to RANKL. Here, we investigated the role of the NLRP3 inflammasome in mouse macrophages, with or without exposure to RANKL. Unexpectedly, we found that NLRP3 expression gradually declined during osteoclastogenesis but could be restored with LPS treatment. LPS and nigericin robustly activated this inflammasome in macrophages, as expected, but not in OCs. Mechanistically, we identified Tmem178, a protein that restrains Ca 2+ release from the endoplasmic reticulum (ER) and highly expressed in OCs, as an inhibitor of this inflammasome. Notably, NLRP3 inflammasome activation was robust in OCs lacking Tmem178 or wild-type (WT) OCs exposed to high calcium concentrations. In vivo studies demonstrated that under the conditions where OCs efficiently release Ca 2+ from bone, inflammasome formation was enhanced. Furthermore, deletion of Nlrp3 rescued osteopenia in Tmem178 −/− mice. Thus, we found that Tmem178 uniquely restricts Ca 2+ release from ER in OCs, thereby suppressing NLRP3 inflammasome activation. One Sentence Summary The NLRP3 inflammasome is silenced in the OC lineage by Tmem178 to prevent pathological bone loss.
Wyatt Tyndall, Nebojsa Kuljic, Michael Thatcher
et al.
Abstract Introduction Patient satisfaction is a critical outcome in total joint arthroplasty (TJA), yet assessing it effectively remains a challenge due to limitations in patient-reported outcome measures (PROMS). While these measures are commonly gathered in clinical settings, additional contact through mail or phone is often needed, and low response rates can affect the validity and reliability of collected data. To improve response rates, this study evaluated various methods of incentivizing patient participation in a randomized trial format, focusing on postal questionnaires. Patients and methods The study investigated three methods to improve response rates: including a gift card with the questionnaire, promising a gift card upon questionnaire completion, and offering no incentive. It also examined whether different monetary values and the inclusion of the surgeon’s name on materials impacted response rates. We tried to determine factors that could improve follow up telephone response rates in the group of patients that failed to return their questionnaires. Results Higher response rates were observed with monetary incentives (P = 0.056), larger amounts of money offered (P = 0.3839) for filling out the questionnaire, and if the surgeon’s details were on the cover letter or questionnaire (P = 0.632). There was no correlation between age and sex and participation. We did find a statistically significant difference in total participation and poorer total knee arthroplasty outcomes scores (P < 0.001). Conclusion Our study supports findings from prior research indicating that monetary incentives and personalized materials can improve response rates, although in this cohort, results were modest. Follow-up calls further boosted response rates, suggesting that multi-modal engagement may be beneficial. Although the response improvements were limited and lacked statistical significance, the study highlights the importance of refining strategies to ensure reliable PROMS data, which is vital for understanding patient outcomes in TJA. Future studies might consider demographic factors and other outreach methods to enhance PROMs data collection.
Orthopedic surgery, Diseases of the musculoskeletal system
Abstract Objectives Takayasu arteritis (TAK) is an inflammatory vasculitis that affects the aorta and its primary branches. The pathogenesis of TAK remains elusive, yet identifying key cell types in the aorta of TAK patients is crucial for uncovering cellular heterogeneity and discovering potential therapeutic targets. Methods This study utilized single-cell transcriptome analysis on aortic specimens from three TAK patients, with control data sourced from a publicly available database (GSE155468). Additionally, bulk RNA sequencing was performed on peripheral CD4 + and CD8 + T cells from eight TAK patients and eight matched healthy volunteers. All participants were recruited at Anzhen Hospital, Capital Medical University, China, between January 2020 and December 2023. Results Single-cell transcriptome analysis identified 11 predominant cell types in aortic tissues, with notable differences in proportions between TAK patients and controls. T cells, B cells, macrophages, smooth muscle cells (SMCs), and fibroblasts exhibited subtype-specific gene expression signatures, with notable changes in interactions between T cells, B cells, and monocyte-macrophages, highlighting their active involvement in the pathogenesis of TAK. Bulk RNA-Seq analysis of peripheral blood T cells from TAK patients showed an upregulation of complement system genes, underscoring the significance of the complement signaling pathway in TAK’s immunopathogenesis. Conclusion The findings underscore the active involvement of various immune and structural cells in the aortic tissues of TAK patients and reveal the presence of the complement signaling pathway in peripheral blood T cells. These insights are instrumental for identifying novel therapeutic targets and developing robust disease monitoring methods for TAK.
Abstract Background The role of osteotomies in deformity correction in congenital pseudarthrosis of the tibia (CPT) remains controversial. This study aims to evaluate the efficacy of tibial closing osteotomy correction for tibial angular deformity in CPT patients. Methods This study selected CPT patients who underwent CPT combined with tibial closing osteotomy in our hospital from January 2011 to May 2022 as the research subjects. The inclusion criterion was children with Crawford IV CPT who also had angular deformities. The PACS system was used to measure the distance between the tibial osteotomy and the tibial pseudarthrosis, as well as the tibial angle. The degree of healing of tibial pseudarthrosis was evaluated via the RUST score. If the RUST score is greater than 8, tibial pseudarthrosis is considered to have achieved primary union. After surgery, X-rays were taken every 2 months until the patient’s tibial pseudarthrosis healed. Results Twenty-three patients with CPT underwent combined surgery and proximal tibial osteotomy. The average age at the time of surgery was 48.7 months (14–158 months). There were 15 males and 8 females, including 17 patients with type 1 neurofibromatosis. Nineteen patients had proximal tibial dysplasia. There were 12 cases on the left and 11 cases on the right. The average angle of the tibia in the preoperative anterior posterior position was 10.3°, and the average angle of the tibia in the preoperative lateral position was 20°. All patients achieved primary union, with an average union time of 4.7 months. The average distance between the osteotomy site and the tibial pseudarthrosis site was 6.5 cm. Twenty-two patients achieved union at the osteotomy site during the healing process of tibial pseudarthrosis. One patient did not achieve union at the osteotomy site, but healing was achieved after a wrapped autogenous iliac bone graft was applied for four months. Conclusion CPT patients with tibial angular deformities can undergo combined surgery and tibial closing osteotomy correction. It has little effect on the healing of the pseudarthrosis of the tibia, and tibial closing osteotomy may be safe and effective for correcting tibial angular deformity. The preliminary findings requiring further validation.
Orthopedic surgery, Diseases of the musculoskeletal system
Plant diseases pose a serious challenge to agriculture by reducing crop yield and affecting food quality. Early detection and classification of these diseases are essential for minimising losses and improving crop management practices. This study applies Convolutional Neural Networks (CNN) and Long Short-Term Memory (LSTM) models to classify plant leaf diseases using a dataset containing 70,295 training images and 17,572 validation images across 38 disease classes. The CNN model was trained using the Adam optimiser with a learning rate of 0.0001 and categorical cross-entropy as the loss function. After 10 training epochs, the model achieved a training accuracy of 99.1% and a validation accuracy of 96.4%. The LSTM model reached a validation accuracy of 93.43%. Performance was evaluated using precision, recall, F1-score, and confusion matrix, confirming the reliability of the CNN-based approach. The results suggest that deep learning models, particularly CNN, enable an effective solution for accurate and scalable plant disease classification, supporting practical applications in agricultural monitoring.
Shih-Huan Huang, Matthew W. Cotton, Tuomas P. J. Knowles
et al.
A central challenge in modeling neurodegenerative diseases is connecting cellular-level mechanisms to tissue-level pathology, in particular to determine whether pathology is driven primarily by cell-autonomous triggers or by propagation from cells that are already in a pathological, runaway aggregation state. To bridge this gap, we here develop a bottom-up physical model that explicitly incorporates these two fundamental cell-level drivers of protein aggregation dynamics. We show that our model naturally explains the characteristic long, slow development of pathology followed by a rapid acceleration, a hallmark of many neurodegenerative diseases. Furthermore, the model reveals the existence of a critical switch point at which the system's dynamics transition from being dominated by slow, spontaneous formation of diseased cells to being driven by fast propagation. This framework provides a robust physical foundation for interpreting pathological data and offers a method to predict which class of therapeutic strategies is best matched to the underlying drivers of a specific disease.
This paper conducts research on the established model and presents the main conclusions . Firstly, by separately considering the infectivity of each of the two infectious diseases and the infectivity of the population simultaneously infected with the two infectious diseases, the existence of three types of boundary equilibrium points is determined, as well as the existence of the interior equilibrium point when the parameters are under specific conditions. Then, the stability of the equilibrium points is analyzed. It is concluded that under different parameter conditions, the stability of the disease free equilibrium point can exhibit various scenarios, such as a stable node or a saddle-node, etc. For the boundary equilibrium points, the situation is more intricate,and a cusp may occur. The stability of the interior equilibrium point under specific conditions is also presented. Finally,the degeneracy of the equilibrium points is studied through the bifurcation theory.Mainly, the saddle-node bifurcation occurring at the interior equilibrium point is obtained, and when the infection rate of the first infectious disease, the infection rate of the second infectious disease, and the infection rate of the co-infected population to other populations are selected as bifurcation parameters, a codimension 3 B-T bifurcation is obtained.
Chronic diseases often progress differently across patients. Rather than randomly varying, there are typically a small number of subtypes for how a disease progresses across patients. To capture this structured heterogeneity, the Subtype and Stage Inference Event-Based Model (SuStaIn) estimates the number of subtypes, the order of disease progression for each subtype, and assigns each patient to a subtype from primarily cross-sectional data. It has been widely applied to uncover the subtypes of many diseases and inform our understanding of them. But how robust is its performance? In this paper, we develop a principled Bayesian subtype variant of the event-based model (BEBMS) and compare its performance to SuStaIn in a variety of synthetic data experiments with varied levels of model misspecification. BEBMS substantially outperforms SuStaIn across ordering, staging, and subtype assignment tasks. Further, we apply BEBMS and SuStaIn to a real-world Alzheimer's data set. We find BEBMS has results that are more consistent with the scientific consensus of Alzheimer's disease progression than SuStaIn.
Plant diseases pose a significant threat to global food security, necessitating accurate and interpretable disease detection methods. This study introduces an interpretable attention-guided Convolutional Neural Network (CNN), CBAM-VGG16, for plant leaf disease detection. By integrating Convolution Block Attention Module (CBAM) at each convolutional stage, the model enhances feature extraction and disease localization. Trained on five diverse plant disease datasets, our approach outperforms recent techniques, achieving high accuracy (up to 98.87%) and demonstrating robust generalization. Here, we show the effectiveness of our method through comprehensive evaluation and interpretability analysis using CBAM attention maps, Grad-CAM, Grad-CAM++, and Layer-wise Relevance Propagation (LRP). This study advances the application of explainable AI in agricultural diagnostics, offering a transparent and reliable system for smart farming. The code of our proposed work is available at https://github.com/BS0111/PlantAttentionCBAM.
Mary K. Horton, Joanne Nititham, Kimberly E. Taylor
et al.
Abstract Background Systemic lupus erythematosus (SLE) has numerous symptoms across organs and an unpredictable flare-remittance pattern. This has made it challenging to understand drivers of long-term SLE outcomes. Our objective was to identify whether changes in DNA methylation over time, in an actively flaring SLE cohort, were associated with remission and whether these changes meaningfully subtype SLE patients. Methods Fifty-nine multi-ethnic SLE patients had clinical visits and DNA methylation profiles at a flare and approximately 3 months later. Methylation was measured using the Illumina EPIC array. We identified sites where methylation change between visits was associated with remission at the follow-up visit using limma package and a time x remission interaction term. Models adjusted for batch, age at diagnosis, time between visits, age at flare, sex, medications, and cell-type proportions. Separately, a paired T-test identified Bonferroni significant methylation sites with ≥ 3% change between visits (n = 546). Methylation changes at these sites were used for unsupervised consensus hierarchical clustering. Associations between clusters and patient features were assessed. Results Nineteen patients fully remitted at the follow-up visit. For 1,953 CpG sites, methylation changed differently for remitters vs. non-remitters (Bonferroni p < 0.05). Nearly half were within genes regulated by interferon. The largest effect was at cg22873177; on average, remitters had 23% decreased methylation between visits while non-remitters had no change. Three SLE patient clusters were identified using methylation differences agnostic of clinical outcomes. All Cluster 1 subjects (n = 12) experienced complete flare remission, despite similar baseline disease activity scores, medications, and demographics as other clusters. Methylation changes at six CpG sites, including within immune-related CD45 and IFI genes, were particularly distinct for each cluster, suggesting these may be good candidates for stratifying patients in the future. Conclusions Changes in DNA methylation during active SLE were associated with remission status and identified subgroups of SLE patients with several distinct clinical and biological characteristics. DNA methylation patterns might help inform SLE subtypes, leading to targeted therapies based on relevant underlying biological pathways.
Abstract Background Pain management for knee osteoarthritis (KOA) patients is challenging. Pain arises from both physiological and psychological interactions, with anxiety and depression potentially contributing as risk factors that hinder effective pain management in KOA patients. Methods Before treatment(T1), A total of 206 elderly inpatients with KOA were enrolled based on initial screening criteria. After treatment (T2), patients were selected based on inclusion and exclusion criteria, and completed follow-up through phone or online questionnaires. The interval between T1 and T2 was three months. Outcome measures included the Visual Analogue Scale (VAS) for pain intensity, Beck Anxiety Inventory (BAI) for anxiety, and Geriatric Depression Scale (GDS) for depression. Descriptive and bivariate analyses were used to evaluate the pain, anxiety and depression of the participants. A cross-lagged model was used to examine the temporal and causal associations among pain, anxiety, and depression. Results 91% of elderly patients with KOA experienced at least mild depression. Furthermore, 31% of patients reported mild or higher levels of anxiety. At the same time, pain, depression, and anxiety were significantly correlated and mutually predictive(all p < 0.01). Across the different time points, Depression and anxiety at T1 positively predicted pain at T2,with correlation coefficients of 0.19 (p < 0.05) and 0.07 (p < 0.05), respectively. Conclusions Anxiety and depression may be potential risk factors limiting the effectiveness of pain management in KOA patients. Clinical treatment should regularly evaluate anxiety and depression levels and integration of psychological interventions or appropriate antianxiety and antidepressant medications. Clinical trial number Not applicable, for the investigative research nature of the study.
R. Maarleveld, H. E. J. Veeger, F. C. T. van der Helm
et al.
Musculoskeletal (MSK) models offer a non-invasive way to understand biomechanical loads on joints and tendons, which are difficult to measure directly. Variations in muscle strength, especially relative differences between muscles, significantly impact model outcomes. Typically, scaled generic MSK models use maximum isometric forces that are not adjusted for different demographics, raising concerns about their accuracy. This review provides an overview on experimentally derived strength parameters, including physiological cross-sectional area (PCSA), muscle mass (Mm), and relative muscle mass (%Mm), which is the relative distribution of muscle mass across the leg. We analysed differences by age and sex, and compared open-source lower limb MSK model parameters with experimental data from 57 studies. Our dataset, with records dating back to 1884, shows that uniformly increasing all maximum isometric forces in MSK models does not capture key muscle ratio differences due to age and sex. Males have a higher proportion of muscle mass in the rectus femoris and semimembranosus muscles, while females have a greater relative muscle mass in the pelvic (gluteus maximus and medius) and ankle muscles (tibialis anterior, tibialis posterior, and extensor digitorum longus). Older adults have a higher relative muscle mass in the gluteus medius, while younger individuals show more in the gastrocnemius. Current MSK models do not accurately represent muscle mass distribution for specific age or sex groups, and none of them accurately reflect female muscle mass distribution. Further research is needed to explore musculotendon age- and sex differences.
Heart diseases rank among the leading causes of global mortality, demonstrating a crucial need for early diagnosis and intervention. Most traditional electrocardiogram (ECG) based automated diagnosis methods are trained at population level, neglecting the customization of personalized ECGs to enhance individual healthcare management. A potential solution to address this limitation is to employ digital twins to simulate symptoms of diseases in real patients. In this paper, we present an innovative prospective learning approach for personalized heart disease detection, which generates digital twins of healthy individuals' anomalous ECGs and enhances the model sensitivity to the personalized symptoms. In our approach, a vector quantized feature separator is proposed to locate and isolate the disease symptom and normal segments in ECG signals with ECG report guidance. Thus, the ECG digital twins can simulate specific heart diseases used to train a personalized heart disease detection model. Experiments demonstrate that our approach not only excels in generating high-fidelity ECG signals but also improves personalized heart disease detection. Moreover, our approach ensures robust privacy protection, safeguarding patient data in model development.
Hiroyuki Sato, Keisuke Suzuki, Atsushi Hashizume
et al.
Progressive cognitive decline spanning across decades is characteristic of Alzheimer's disease (AD). Various predictive models have been designed to realize its early onset and study the long-term trajectories of cognitive test scores across populations of interest. Research efforts have been geared towards superimposing patients' cognitive test scores with the long-term trajectory denoting gradual cognitive decline, while considering the heterogeneity of AD. Multiple trajectories representing cognitive assessment for the long-term have been developed based on various parameters, highlighting the importance of classifying several groups based on disease progression patterns. In this study, a novel method capable of self-organized prediction, classification, and the overlay of long-term cognitive trajectories based on short-term individual data was developed, based on statistical and differential equation modeling. We validated the predictive accuracy of the proposed method for the long-term trajectory of cognitive test score results on two cohorts: the Alzheimer's Disease Neuroimaging Initiative (ADNI) study and the Japanese ADNI study. We also presented two practical illustrations of the simultaneous evaluation of risk factor associated with both the onset and the longitudinal progression of AD, and an innovative randomized controlled trial design for AD that standardizes the heterogeneity of patients enrolled in a clinical trial. These resources would improve the power of statistical hypothesis testing and help evaluate the therapeutic effect. The application of predicting the trajectory of longitudinal disease progression goes beyond AD, and is especially relevant for progressive and neurodegenerative disorders.
With the widespread application of deep learning technology in medical image analysis, the effective explanation of model predictions and improvement of diagnostic accuracy have become urgent problems that need to be solved. Attribution methods have become key tools to help doctors better understand the diagnostic basis of models, and are used to explain and localize diseases in medical images. However, previous methods suffer from inaccurate and incomplete localization problems for fundus diseases with complex and diverse structures. To solve these problems, we propose a weakly supervised interpretable fundus disease localization method called hierarchical salient patch identification (HSPI) that can achieve interpretable disease localization using only image-level labels and a neural network classifier (NNC). First, we propose salient patch identification (SPI), which divides the image into several patches and optimizes consistency loss to identify which patch in the input image is most important for the network's prediction, in order to locate the disease. Second, we propose a hierarchical identification strategy to force SPI to analyze the importance of different areas to neural network classifier's prediction to comprehensively locate disease areas. Conditional peak focusing is then introduced to ensure that the mask vector can accurately locate the disease area. Finally, we propose patch selection based on multi-sized intersections to filter out incorrectly or additionally identified non-disease regions. We conduct disease localization experiments on fundus image datasets and achieve the best performance on multiple evaluation metrics compared to previous interpretable attribution methods. Additional ablation studies are conducted to verify the effectiveness of each method.