Internal Test Set Research Articles

Deep learning-based detection of primary bone tumors around the knee joint on radiographs: a multicenter study.

Most primary bone tumors are often found in the bone around the knee joint. However, the detection of primary bone tumors on radiographs can be challenging for the inexperienced or junior radiologist. This study aimed to develop a deep learning (DL) model for the detection of primary bone tumors around the knee joint on radiographs. From four tertiary referral centers, we recruited 687 patients diagnosed with bone tumors (including osteosarcoma, chondrosarcoma, giant cell tumor of bone, bone cyst, enchondroma, fibrous dysplasia, etc.; 417 males, 270 females; mean age 22.8±13.2 years) by postoperative pathology or clinical imaging/follow-up, and 1,988 participants with normal bone radiographs (1,152 males, 836 females; mean age 27.9±12.2 years). The dataset was split into a training set for model development, an internal independent and an external test set for model validation. The trained model located bone tumor lesions and then detected tumor patients. Receiver operating characteristic curves and Cohen's kappa coefficient were used for evaluating detection performance. We compared the model's detection performance with that of two junior radiologists in the internal test set using permutation tests. The DL model correctly localized 94.5% and 92.9% bone tumors on radiographs in the internal and external test set, respectively. An accuracy of 0.964/0.920, and an area under the receiver operating characteristic curve (AUC) of 0.981/0.990 in DL detection of bone tumor patients were for the internal and external test set, respectively. Cohen's kappa coefficient of the model in the internal test set was significantly higher than that of the two junior radiologists with 4 and 3 years of experience in musculoskeletal radiology (Model vs. Reader A, 0.927 vs. 0.777, P<0.001; Model vs. Reader B, 0.927 vs. 0.841, P=0.033). The DL model achieved good performance in detecting primary bone tumors around the knee joint. This model had better performance than those of junior radiologists, indicating the potential for the detection of bone tumors on radiographs.

Radiomics Analysis of CTO Plaques for Predicting Successful Guidewire Crossing Within 30 Minutes of PCI.

Coronary computed tomography angiography provides valuable information for evaluating the difficulty of chronic total occlusion (CTO) percutaneous coronary intervention. This study aimed to investigate the value of CTO plaque characteristics derived from radiomics analysis for predicting the difficulty of percutaneous coronary intervention. Patients with CTO were retrospectively enrolled from a hospital as training and internal test sets and from the other 2 territory hospitals as external test sets. Radiomics characteristics were extracted from the CTO segment on coronary computed tomography angiography. Radiomics and combined models were developed to predict successful guidewire crossing within 30 minutes (guidewire success) of CTO percutaneous coronary intervention. Subgroup analysis was conducted to investigate the influence of potential risk factors on the radiomics model performance. A total of 551 patients (median, 60; interquartile range, 52.00-66.00 years, 460 men) with 565 CTO lesions were finally enrolled. In the training, internal test, and external test sets, 203 of 357, 85 of 149, and 38 of 59 CTO lesions achieved guidewire success, respectively. Six radiomics features were selected for constructing the radiomics model. In the external test set, the area under the receiver operating characteristic curve of the radiomics model was significantly higher than prior prediction models (P<0.05 for all) with the area under the receiver operating characteristic curve, accuracy, sensitivity, and specificity of 0.86, 74.58%, 81.58%, and 61.90%, respectively. The performance of the radiomics model was dependent on calcification, CTO location, adjacent branch(es), and operator caseload. CTO characteristics revealed by radiomics analysis can be used as effective imaging biomarkers for predicting guidewire success. However, the performance of the radiomics model depends on anatomic and operator factors.

Artificial intelligence improves the diagnosis of human leukocyte antigen (HLA)-B27-negative axial spondyloarthritis based on multi-sequence magnetic resonance imaging and clinical features.

Axial spondyloarthritis (axSpA) is frequently diagnosed late, particularly in human leukocyte antigen (HLA)-B27-negative patients, resulting in a missed opportunity for optimal treatment. This study aimed to develop an artificial intelligence (AI) tool, termed NegSpA-AI, using sacroiliac joint (SIJ) magnetic resonance imaging (MRI) and clinical SpA features to improve the diagnosis of axSpA in HLA-B27-negative patients. We retrospectively included 454 HLA-B27-negative patients with rheumatologist-diagnosed axSpA or other diseases (non-axSpA) from the Third Affiliated Hospital of Southern Medical University and Nanhai Hospital between January 2010 and August 2021. They were divided into a training set (n=328) for 5-fold cross-validation, an internal test set (n=72), and an independent external test set (n=54). To construct a prospective test set, we further enrolled 87 patients between September 2021 and August 2023 from the Third Affiliated Hospital of Southern Medical University. MRI techniques employed included T1-weighted (T1W), T2-weighted (T2W), and fat-suppressed (FS) sequences. We developed NegSpA-AI using a deep learning (DL) network to differentiate between axSpA and non-axSpA at admission. Furthermore, we conducted a reader study involving 4 radiologists and 2 rheumatologists to evaluate and compare the performance of independent and AI-assisted clinicians. NegSpA-AI demonstrated superior performance compared to the independent junior rheumatologist (≤5 years of experience), achieving areas under the curve (AUCs) of 0.878 [95% confidence interval (CI): 0.786-0.971], 0.870 (95% CI: 0.771-0.970), and 0.815 (95% CI: 0.714-0.915) on the internal, external, and prospective test sets, respectively. The assistance of NegSpA-AI promoted discriminating accuracy, sensitivity, and specificity of independent junior radiologists by 7.4-11.5%, 1.0-13.3%, and 7.4-20.6% across the 3 test sets (all P<0.05). On the prospective test set, AI assistance also improved the diagnostic accuracy, sensitivity, and specificity of independent junior rheumatologists by 7.7%, 7.7%, and 6.9%, respectively (all P<0.01). The proposed NegSpA-AI effectively improves radiologists' interpretations of SIJ MRI and rheumatologists' diagnoses of HLA-B27-negative axSpA.

Fully Automated Deep Learning Model to Detect Clinically Significant Prostate Cancer at MRI.

Background Multiparametric MRI can help identify clinically significant prostate cancer (csPCa) (Gleason score ≥7) but is limited by reader experience and interobserver variability. In contrast, deep learning (DL) produces deterministic outputs. Purpose To develop a DL model to predict the presence of csPCa by using patient-level labels without information about tumor location and to compare its performance with that of radiologists. Materials and Methods Data from patients without known csPCa who underwent MRI from January 2017 to December 2019 at one of multiple sites of a single academic institution were retrospectively reviewed. A convolutional neural network was trained to predict csPCa from T2-weighted images, diffusion-weighted images, apparent diffusion coefficient maps, and T1-weighted contrast-enhanced images. The reference standard was pathologic diagnosis. Radiologist performance was evaluated as follows: Radiology reports were used for the internal test set, and four radiologists' PI-RADS ratings were used for the external (ProstateX) test set. The performance was compared using areas under the receiver operating characteristic curves (AUCs) and the DeLong test. Gradient-weighted class activation maps (Grad-CAMs) were used to show tumor localization. Results Among 5735 examinations in 5215 patients (mean age, 66 years ± 8 [SD]; all male), 1514 examinations (1454 patients) showed csPCa. In the internal test set (400 examinations), the AUC was 0.89 and 0.89 for the DL classifier and radiologists, respectively (P = .88). In the external test set (204 examinations), the AUC was 0.86 and 0.84 for the DL classifier and radiologists, respectively (P = .68). DL classifier plus radiologists had an AUC of 0.89 (P < .001). Grad-CAMs demonstrated activation over the csPCa lesion in 35 of 38 and 56 of 58 true-positive examinations in internal and external test sets, respectively. Conclusion The performance of a DL model was not different from that of radiologists in the detection of csPCa at MRI, and Grad-CAMs localized the tumor. © RSNA, 2024 Supplemental material is available for this article. See also the editorial by Johnson and Chandarana in this issue.

Development and Validation of a CT-Radiomics Nomogram for the Diagnosis of Small Prevascular Mediastinal Nodules: Reducing Nontherapeutic Surgeries

Rationale and ObjectivesThe preoperative diagnosis of small prevascular mediastinal nodules (SPMNs) presents a challenge, often leading to unnecessary surgical interventions. Our objective was to develop a nomogram based on preoperative CT radiomics features, serving as a non-invasive diagnostic tool for SPMNs. Materials and MethodsPatients with surgically resected SPMNs from two medical centers between January 2018 and December 2022 were retrospectively reviewed. Radiomics features were extracted and screened from preoperative CT images. Logistic regression was employed to establish clinical, radiomics, and hybrid models for differentiating thymic epithelial tumors (TETs) from cysts. The performance of these models was validated in both internal and external test sets by area under the receiver operating characteristic curve (AUC), while also comparing their diagnostic capability with human experts. ResultsThe study enrolled a total of 363 patients (median age, 53 years [IQR:45-59 years]; 175 [48.2%] males) for model development and validation, including 136 TETs and 227 cysts. Lesions' enhancement status, shape, calcification, and rad-score were identified as independent factors for distinction. The hybrid model demonstrated superior diagnostic performance compared to other models and human experts, with an AUC of 0.95 (95% CI:0.92-0.98), 0.94 (95% CI:0.89-0.99), and 0.93 (95% CI:0.83-1.00) in the training set, internal test set, and external test set respectively. The calibration curve of the model demonstrated excellent fit, while decision curve analysis underscored its clinical value. ConclusionThe radiomics-based nomogram effectively discriminates between the most prevalent types of SPMNs, namely TETs and cysts, thus presenting a promising tool for treatment guidance.

Knowledge-Augmented Deep Learning for Segmenting and Detecting Cerebral Aneurysms With CT Angiography: A Multicenter Study.

Background Deep learning (DL) could improve the labor-intensive, challenging processes of diagnosing cerebral aneurysms but requires large multicenter data sets. Purpose To construct a DL model using a multicenter data set for accurate cerebral aneurysm segmentation and detection on CT angiography (CTA) images and to compare its performance with radiology reports. Materials and Methods Consecutive head or head and neck CTA images of suspected unruptured cerebral aneurysms were gathered retrospectively from eight hospitals between February 2018 and October 2021 for model development. An external test set with reference standard digital subtraction angiography (DSA) scans was obtained retrospectively from one of the eight hospitals between February 2022 and February 2023. Radiologists (reference standard) assessed aneurysm segmentation, while model performance was evaluated using the Dice similarity coefficient (DSC). The model's aneurysm detection performance was assessed by sensitivity and comparing areas under the receiver operating characteristic curves (AUCs) between the model and radiology reports in the DSA data set with use of the DeLong test. Results Images from 6060 patients (mean age, 56 years ± 12 [SD]; 3375 [55.7%] female) were included for model development (training: 4342; validation: 1086; and internal test set: 632). Another 118 patients (mean age, 59 years ± 14; 79 [66.9%] female) were included in an external test set to evaluate performance based on DSA. The model achieved a DSC of 0.87 for aneurysm segmentation performance in the internal test set. Using DSA, the model achieved 85.7% (108 of 126 aneurysms [95% CI: 78.1, 90.1]) sensitivity in detecting aneurysms on per-vessel analysis, with no evidence of a difference versus radiology reports (AUC, 0.93 [95% CI: 0.90, 0.95] vs 0.91 [95% CI: 0.87, 0.94]; P = .67). Model processing time from reconstruction to detection was 1.76 minutes ± 0.32 per scan. Conclusion The proposed DL model could accurately segment and detect cerebral aneurysms at CTA with no evidence of a significant difference in diagnostic performance compared with radiology reports. © RSNA, 2024 Supplemental material is available for this article. See also the editorial by Payabvash in this issue.

Fully automated hybrid approach on conventional MRI for triaging clinically significant liver fibrosis: A multi-center cohort study.

Establishing reliable noninvasive tools to precisely diagnose clinically significant liver fibrosis (SF, ≥F2) remains an unmet need. We aimed to build a combined radiomics-clinic (CoRC) model for triaging SF and explore the additive value of the CoRC model to transient elastography-based liver stiffness measurement (FibroScan, TE-LSM). This retrospective study recruited 595 patients with biopsy-proven liver fibrosis at two centers between January 2015 and December 2021. At Center 1, the patients before December 2018 were randomly split into training (276) and internal test (118) sets, the remaining were time-independent as a temporal test set (96). Another data set (105) from Center 2 was collected for external testing. Radiomics scores were built with selected features from Deep learning-based (ResUNet) automated whole liver segmentations on MRI (T2FS and delayed enhanced-T1WI). The CoRC model incorporated radiomics scores and relevant clinical variables with logistic regression, comparing routine approaches. Diagnostic performance was evaluated by the area under the receiver operating characteristic curve (AUC). The additive value of the CoRC model to TE-LSM was investigated, considering necroinflammation. The CoRC model achieved AUCs of 0.79 (0.70, 0.86), 0.82 (0.73, 0.89), and 0.81 (0.72-0.91), outperformed FIB-4, APRI (all p < 0.05) in the internal, temporal, and external test sets and maintained the discriminatory power in G0-1 subgroups (AUCs range, 0.85-0.86; all p < 0.05). The AUCs of joint CoRC-LSM model were 0.86 (0.79-0.94), and 0.81 (0.72-0.90) in the internal and temporal sets (p = 0.01). The CoRC model was useful for triaging SF, and may add value to TE-LSM.

Machine Learning Based on Clinical Information and Integrated CT Radiomics to Predict Local Recurrence of Stage Ia Lung Adenocarcinoma after Microwave Ablation

PurposeTo develop and compare 3 different machine learning–based models of clinical information and integrated radiomics features predicting the local recurrence of Stage Ia lung adenocarcinoma after microwave ablation (MWA) for assisting clinical decision making. Materials and MethodsThe data of 360 patients with Stage Ia lung adenocarcinoma who underwent MWA were included in the training (n = 208), internal test (n = 90), and external test (n = 62) sets based on the inclusion and exclusion criteria. The predictors associated with local recurrence were identified using univariate and multivariate analyses of clinical information. The integrated radiomics features were extracted from pre-MWA and post-MWA (scanned immediately after the ablation) computed tomography (CT) images, and 10 radiomics features were selected by the t-test and least absolute shrinkage and selection operator. The L2-logistic regression of machine learning was applied for the clinical model, CT radiomics model, and combined model including clinical predictors and radiomics features. Model performance was evaluated by the receiver operating characteristic and decision curve analysis. ResultsThe ablative margin was an independent clinical predictor (P = 0.001; odds ratio [OR], 0.46; 95% CI, 0.29–0.73). The combined model showed the highest area under the curve value among the 3 models (training, 0.86; 95% CI, 0.81–0.91; internal test, 0.93; 95% CI, 0.87–0.98; external test, 0.89; 95% CI, 0.79–0.96). ConclusionsThe combined model could accurately predict the local recurrence of Stage Ia lung adenocarcinoma after MWA to better support a clinical decision.

Early Warning Models Using Machine Learning to Predict Sepsis-Associated Chronic Critical Illness: A Study Based on the Medical Information Mart for Intensive Care Database.

Background Patients with chronic critical illness (CCI) experience poor prognoses and incur high medical costs. However, there is currently limited clinical awareness of sepsis-associated CCI, resulting in insufficient vigilance.Therefore, it is necessary to build a machine learning model that can predict whether sepsis patients will developCCI. Methods Clinical data on19,077 sepsis patients from the Medical Information Mart for Intensive Care IV (MIMIC-IV) database were analyzed. Predictive factors were identified using the Student's t-test, Mann-Whitney Utest, or χ 2test. Six machine learning classification models, namely, the logistic regression, support vector machine, decision tree, random forest, extreme gradient enhancement, and artificial neural network, were established. The optimal model was selected on the basis of its performance. Calibration curves wereused to evaluate the accuracy of model classification, while the external validation dataset was used to evaluate the performance of the model. Results Thirty-seven characteristics,such as elevated alanine aminotransferase, rapid heart rate, and high Logistic Organ Dysfunction System scores, were identified as risk factors for developingCCI. The area under the receiver operating characteristic curve (AUROC) values forall models were above 0.73 onthe internal test set. Among them, the extreme gradient enhancement model exhibited superior performance (F1score=0.91, AUROC=0.91, Brierscore=0.052). It alsoexhibited stable prediction performance onthe external validation set (AUROC =0.72). Conclusion A machine learning model was established to predict whether sepsis patients will develop CCI. It can provide useful predictive information for clinical decision-making.

Graph transformer based transfer learning for aqueous pKa prediction of organic small molecules

The acid-base dissociation constant (pKa) is an essential physicochemical parameter that indicates the extent of proton dissociation. However, accurately predicting pKa values is still challenging due to limited data availability for organic small molecules in aqueous solutions. In this work, we propose an open-source pKa prediction tool based on Graph Transformer, which combines the graph neural network and transformer to take both local and global information into account. To address the limitation of data scarcity, the experimental dataset is expanded, while transfer learning is employed by pre-training on the ChEMBL (computational) dataset and fine-tuning on the experimental dataset. The performance and generalization of the obtained models are comprehensively evaluated on both internal and external test sets. Additionally, two exemplary applications, namely the screening of tertiary amines for CO2 absorption and acidic ionic liquids for reactive extraction, are investigated to validate the reliability of our tool. The results show that our model can effectively guide the identification of promising candidates for the target applications by achieving accurate pKa predictions.

Prognostic and therapeutic value of the Eph/Ephrin signaling pathway in pancreatic cancer explored based on bioinformatics

Pancreatic cancer (PC) is one of the most common malignant tumors of the digestive tract and has a very high mortality rate worldwide. Different PC patients may respond differently to therapy and develop therapeutic resistance due to the complexity and variety of the tumor microenvironment. The Eph/ephrin signaling pathway is extensively involved in tumor-related biological functions. However, the key function of the Eph/ephrin signaling pathway in PC has not been fully elucidated. We first explored a pan-cancer overview of Eph/ephrin signaling pathway genes (EPGs). Then we grouped the PC patients into 3 subgroups based on EPG expression levels. Significantly different prognoses and tumor immune microenvironments between different subtypes further validate Eph/ephrin's important role in the pathophysiology of PC. Additionally, we estimated the IC50 values for several commonly used molecularly targeted drugs used to treat PC in the three clusters, which could help patients receive a more personalized treatment plan. Following a progressive screening of optimal genes, we established a prognostic signature and validated it in internal and external test sets. The receiver operating characteristic (ROC) curves of our model exhibited great predictive performance. Meanwhile, we further validated the results through qRT-PCR and immunohistochemistry. Overall, this research provides fresh clues on the prognosis and therapy of PC as well as the theoretical groundwork for future Eph/ephrin signaling pathway research.

Deep learning-based fully automatic Risser stage assessment model using abdominal radiographs.

Artificial intelligence has been increasingly used in medical imaging and has demonstrated expert level performance in image classification tasks. To develop a fully automatic approach for determining the Risser stage using deep learning on abdominal radiographs. In this multicenter study, 1,681 supine abdominal radiographs (age range, 9-18years, 50% female) obtained between January 2019 and April 2022 were collected retrospectively from three medical institutions and graded manually using the United States Risser staging system. A total of 1,577 images from Hospitals 1 and 2 were used for development, and 104 images from Hospital 3 for external validation. From each radiograph, right and left iliac crest patch images were extracted using the pelvic bone segmentation model DeepLabv3 + with the EfficientNet-B0 encoder trained with 90 digitally reconstructed radiographs from pelvic computed tomography scans with a pelvic bone mask. Using these patch images, ConvNeXt-B was trained to grade according to the Risser classification. The model's performance was evaluated using accuracy, area under the receiver operating characteristic curve (AUROC), and mean absolute error. The fully automatic Risser stage assessment model showed an accuracy of 0.87 and 0.75, mean absolute error of 0.13 and 0.26, and AUROC of 0.99 and 0.95 on internal and external test sets, respectively. We developed a deep learning-based, fully automatic segmentation and classification model for Risser stage assessment using abdominal radiographs.

Second opinion machine learning for fast-track pathway assignment in hip and knee replacement surgery: the use of patient-reported outcome measures

BackgroundThe frequency of hip and knee arthroplasty surgeries has been rising steadily in recent decades. This trend is attributed to an aging population, leading to increased demands on healthcare systems. Fast Track (FT) surgical protocols, perioperative procedures designed to expedite patient recovery and early mobilization, have demonstrated efficacy in reducing hospital stays, convalescence periods, and associated costs. However, the criteria for selecting patients for FT procedures have not fully capitalized on the available patient data, including patient-reported outcome measures (PROMs).MethodsOur study focused on developing machine learning (ML) models to support decision making in assigning patients to FT procedures, utilizing data from patients’ self-reported health status. These models are specifically designed to predict the potential health status improvement in patients initially selected for FT. Our approach focused on techniques inspired by the concept of controllable AI. This includes eXplainable AI (XAI), which aims to make the model’s recommendations comprehensible to clinicians, and cautious prediction, a method used to alert clinicians about potential control losses, thereby enhancing the models’ trustworthiness and reliability.ResultsOur models were trained and tested using a dataset comprising 899 records from individual patients admitted to the FT program at IRCCS Ospedale Galeazzi-Sant’Ambrogio. After training and selecting hyper-parameters, the models were assessed using a separate internal test set. The interpretable models demonstrated performance on par or even better than the most effective ‘black-box’ model (Random Forest). These models achieved sensitivity, specificity, and positive predictive value (PPV) exceeding 70%, with an area under the curve (AUC) greater than 80%. The cautious prediction models exhibited enhanced performance while maintaining satisfactory coverage (over 50%). Further, when externally validated on a separate cohort from the same hospital-comprising patients from a subsequent time period-the models showed no pragmatically notable decline in performance.ConclusionsOur results demonstrate the effectiveness of utilizing PROMs as basis to develop ML models for planning assignments to FT procedures. Notably, the application of controllable AI techniques, particularly those based on XAI and cautious prediction, emerges as a promising approach. These techniques provide reliable and interpretable support, essential for informed decision-making in clinical processes.

Deep learning model based on endoscopic images predicting treatment response in locally advanced rectal cancer undergo neoadjuvant chemoradiotherapy: a multicenter study

PurposeNeoadjuvant chemoradiotherapy has been the standard practice for patients with locally advanced rectal cancer. However, the treatment response varies greatly among individuals, how to select the optimal candidates for neoadjuvant chemoradiotherapy is crucial. This study aimed to develop an endoscopic image-based deep learning model for predicting the response to neoadjuvant chemoradiotherapy in locally advanced rectal cancer.MethodsIn this multicenter observational study, pre-treatment endoscopic images of patients from two Chinese medical centers were retrospectively obtained and a deep learning-based tumor regression model was constructed. Treatment response was evaluated based on the tumor regression grade and was defined as good response and non-good response. The prediction performance of the deep learning model was evaluated in the internal and external test sets. The main outcome was the accuracy of the treatment prediction model, measured by the AUC and accuracy.ResultsThis deep learning model achieved favorable prediction performance. In the internal test set, the AUC and accuracy were 0.867 (95% CI: 0.847–0.941) and 0.836 (95% CI: 0.818–0.896), respectively. The prediction performance was fully validated in the external test set, and the model had an AUC of 0.758 (95% CI: 0.724–0.834) and an accuracy of 0.807 (95% CI: 0.774–0.843).ConclusionThe deep learning model based on endoscopic images demonstrated exceptional predictive power for neoadjuvant treatment response, highlighting its potential for guiding personalized therapy.

Uveal melanoma distant metastasis prediction system: A retrospective observational study based on machine learning.

Uveal melanoma (UM) patients face a significant risk of distant metastasis, closely tied to a poor prognosis. Despite this, there is a dearth of research utilizing big data to predict UM distant metastasis. This study leveraged machine learning methods on the Surveillance, Epidemiology, and End Results (SEER) database to forecast the risk probability of distant metastasis. Therefore, the information on UM patients from the SEER database (2000-2020) was split into a 7:3 ratio training set and an internal test set based on distant metastasis presence. Univariate and multivariate logistic regression analyses assessed distant metastasis risk factors. Six machine learning methods constructed a predictive model post-feature variable selection. The model evaluation identified the multilayer perceptron (MLP) as optimal. Shapley additive explanations (SHAP) interpreted the chosen model. A web-based calculator personalized risk probabilities for UM patients. The results show that nine feature variables contributed to the machine learning model. The MLP model demonstrated superior predictive accuracy (Precision = 0.788; ROC AUC = 0.876; PR AUC = 0.788). Grade recode, age, primary site, time from diagnosis to treatment initiation, and total number of malignant tumors were identified as distant metastasis risk factors. Diagnostic method, laterality, rural-urban continuum code, and radiation recode emerged as protective factors. The developed web calculator utilizes the MLP model for personalized risk assessments. In conclusion, the MLP machine learning model emerges as the optimal tool for predicting distant metastasis in UM patients. This model facilitates personalized risk assessments, empowering early and tailored treatment strategies.

Multimodal Machine Learning-Based Marker Enables Early Detection and Prognosis Prediction for Hyperuricemia.

Hyperuricemia (HUA) has emerged as the second most prevalent metabolic disorder characterized by prolonged and asymptomatic period, triggering gout and metabolism-related outcomes. Early detection and prognosis prediction for HUA and gout are crucial for pre-emptive interventions. Integrating genetic and clinical data from 421287 UK Biobank and 8900 Nanfang Hospital participants, a stacked multimodal machine learning model is developed and validated to synthesize its probabilities as an in-silico quantitative marker for hyperuricemia (ISHUA). The model demonstrates satisfactory performance in detecting HUA, exhibiting area under the curves (AUCs) of 0.859, 0.836, and 0.779 within the train, internal, and external test sets, respectively. ISHUA is significantly associated with gout and metabolism-related outcomes, effectively classifying individuals into low- and high-risk groups for gout in the train (AUC, 0.815) and internal test (AUC, 0.814) sets. The high-risk group shows increased susceptibility to metabolism-related outcomes, and participants with intermediate or favorable lifestyle profiles have hazard ratios of 0.75 and 0.53 for gout compared with those with unfavorable lifestyles. Similar trends are observed for other metabolism-related outcomes. The multimodal machine learning-based ISHUA marker enables personalized risk stratification for gout and metabolism-related outcomes, and it is unveiled that lifestyle changes can ameliorate these outcomes within high-risk group, providing guidance for preventive interventions.

Efficient deep learning-based automated diagnosis from echocardiography with contrastive self-supervised learning

BackgroundAdvances in self-supervised learning (SSL) have enabled state-of-the-art automated medical image diagnosis from small, labeled datasets. This label efficiency is often desirable, given the difficulty of obtaining expert labels for medical image recognition tasks. However, most efforts toward SSL in medical imaging are not adapted to video-based modalities, such as echocardiography.MethodsWe developed a self-supervised contrastive learning approach, EchoCLR, for echocardiogram videos with the goal of learning strong representations for efficient fine-tuning on downstream cardiac disease diagnosis. EchoCLR pretraining involves (i) contrastive learning, where the model is trained to identify distinct videos of the same patient, and (ii) frame reordering, where the model is trained to predict the correct of video frames after being randomly shuffled.ResultsWhen fine-tuned on small portions of labeled data, EchoCLR pretraining significantly improves classification performance for left ventricular hypertrophy (LVH) and aortic stenosis (AS) over other transfer learning and SSL approaches across internal and external test sets. When fine-tuning on 10% of available training data (519 studies), an EchoCLR-pretrained model achieves 0.72 AUROC (95% CI: [0.69, 0.75]) on LVH classification, compared to 0.61 AUROC (95% CI: [0.57, 0.64]) with a standard transfer learning approach. Similarly, using 1% of available training data (53 studies), EchoCLR pretraining achieves 0.82 AUROC (95% CI: [0.79, 0.84]) on severe AS classification, compared to 0.61 AUROC (95% CI: [0.58, 0.65]) with transfer learning.ConclusionsEchoCLR is unique in its ability to learn representations of echocardiogram videos and demonstrates that SSL can enable label-efficient disease classification from small amounts of labeled data.

Predicting the Prognosis of HIFU Ablation of Uterine Fibroids Using a Deep Learning-Based 3D Super-Resolution DWI Radiomics Model: A Multicenter Study

Rationale and ObjectivesTo assess the feasibility and efficacy of a deep learning based three-dimensional (3D) super-resolution diffusion-weighted imaging (DWI) radiomics model in predicting the prognosis of high-intensity focused ultrasound (HIFU) ablation of uterine fibroids. MethodsThis retrospective study included 360 patients with uterine fibroids who received HIFU treatment, including Center A (training set: N=240; internal testing set: N=60) and Center B (external testing set: N=60) and were classified as having a favorable or unfavorable prognosis based on the postoperative non-perfusion volume ratio. A deep transfer learning approach was used to construct super-resolution DWI (SR-DWI) based on conventional high-resolution DWI (HR-DWI), and 1198 radiomics features were extracted from manually segmented regions of interest in both image types. Following data preprocessing and feature selection, radiomics models were constructed for HR-DWI and SR-DWI using Support Vector Machine (SVM), Random Forest (RF), and Light Gradient Boosting Machine (LightGBM) algorithms, with performance evaluated using area under the curve (AUC) and decision curves. ResultAll DWI radiomics models demonstrated superior AUC in predicting HIFU ablated uterine fibroids prognosis compared to expert radiologists (AUC: 0.706, 95% CI: 0.647-0.748). When utilizing different machine learning algorithms, the HR-DWI model achieved AUC values of 0.805 (95% CI: 0.679-0.931) with SVM, 0.797 (95% CI: 0.672-0.921) with RF, and 0.770 (95% CI: 0.631-0.908) with LightGBM. Meanwhile, the SR-DWI model outperformed the HR-DWI model (P<0.05) across all algorithms, with AUC values of 0.868 (95% CI: 0.775-0.960) with SVM, 0.824 (95% CI: 0.715-0.934) with RF, and 0.821 (95% CI: 0.709-0.933) with LightGBM. And decision curve analysis further confirmed the good clinical value of the models. ConclusionDeep learning based 3D SR-DWI radiomics model demonstrated favorable feasibility and effectiveness in predicting the prognosis of HIFU ablated uterine fibroids, which was superior to HR-DWI model and assessment by expert radiologists.

Development and validation of a deep learning algorithm for the prediction of serum creatinine in critically ill patients

Abstract Objectives Serum creatinine (SCr) is the primary biomarker for assessing kidney function; however, it may lag behind true kidney function, especially in instances of acute kidney injury (AKI). The objective of the work is to develop Nephrocast, a deep-learning model to predict next-day SCr in adult patients treated in the intensive care unit (ICU). Materials and Methods Nephrocast was trained and validated, temporally and prospectively, using electronic health record data of adult patients admitted to the ICU in the University of California San Diego Health (UCSDH) between January 1, 2016 and June 22, 2024. The model features consisted of demographics, comorbidities, vital signs and laboratory measurements, and medications. Model performance was evaluated by mean absolute error (MAE) and root-mean-square error (RMSE) and compared against the prediction day’s SCr as a reference. Results A total of 28 191 encounters met the eligibility criteria, corresponding to 105 718 patient-days. The median (interquartile range [IQR]) MAE and RMSE in the internal test set were 0.09 (0.085-0.09) mg/dL and 0.15 (0.146-0.152) mg/dL, respectively. In the prospective validation, the MAE and RMSE were 0.09 mg/dL and 0.14 mg/dL, respectively. The model’s performance was superior to the reference SCr. Discussion and Conclusion Our model demonstrated good performance in predicting next-day SCr by leveraging clinical data routinely collected in the ICU. The model could aid clinicians in in identifying high-risk patients for AKI, predicting AKI trajectory, and informing the dosing of renally eliminated drugs.

A Comparison of Variable Input Strategies used for Risk-adjustment Models of Antimicrobial Use

Background: External comparisons of antimicrobial use (AU) may be more informative if adjusted for encounter characteristics. Optimal methods to define input variables for encounter-level risk-adjustment models of AU are not established. Methods: This retrospective analysis of electronic health record data included 50 US hospitals in 2020-2021. We used NHSN definitions for all antibacterials days of therapy (DOT), including adult and pediatric encounters with at least 1 day present in inpatient locations. We assessed 4 methods to define input variables: 1) diagnosis-related group (DRG) categories by Yu et al., 2) adjudicated Elixhauser comorbidity categories by Goodman et al., 3) all Clinical Classification Software Refined (CCSR) diagnosis and procedure categories, and 4) adjudicated CCSR categories where codes not appropriate for AU risk-adjustment were excluded by expert consensus, requiring review of 867 codes over 4 months to attain consensus. Data were split randomly, stratified by bed size as follows: 1) training dataset including two-thirds of encounters among two-thirds of hospitals; 2) internal testing set including one-third of encounters within training hospitals, and 3) external testing set including the remaining one-third of hospitals. We used a gradient-boosted machine (GBM) tree-based model and two-staged approach to first identify encounters with zero DOT, then estimate DOT among those with &gt;0.5 probability of receiving antibiotics. Accuracy was assessed using mean absolute error (MAE) in testing datasets. Correlation plots compared model estimates and observed DOT among testing datasets. The top 20 most influential variables were defined using modeled variable importance. Results: Our datasets included 629,445 training, 314,971 internal testing, and 419,109 external testing encounters. Demographic data included 41% male, 59% non-Hispanic White, 25% non-Hispanic Black, 9% Hispanic, and 5% pediatric encounters. DRG was missing in 29% of encounters. MAE was lower in pediatrics as compared to adults, and lowest for models incorporating CCSR inputs (Figure 1). Performance in internal and external testing was similar, though Goodman/Elixhauser variable strategies were less accurate in external testing and underestimated long DOT outliers (Figure 2). Agnostic and adjudicated CCSR model estimates were highly correlated; their influential variables lists were similar (Figure 3). Conclusion: Larger numbers of CCSR diagnosis and procedure inputs improved risk-adjustment model accuracy compared with prior strategies. Variable importance and accuracy were similar for agnostic and adjudicated approaches. However, maintaining adjudications by experts would require significant time and potentially introduce personal bias. If findings are confirmed, the need for expert adjudication of input variables should be reconsidered.Disclosure: Elizabeth Dodds Ashley: Advisor- HealthTrackRx. David J Weber: Consultant on vaccines: Pfizer; DSMB chair: GSK; Consultant on disinfection: BD, GAMA, PDI, Germitec