External Test Set Research Articles

Integrated Deep Learning Model for the Detection, Segmentation, and Morphologic Analysis of Intracranial Aneurysms Using CT Angiography.

"Just Accepted" papers have undergone full peer review and have been accepted for publication in Radiology: Artificial Intelligence. This article will undergo copyediting, layout, and proof review before it is published in its final version. Please note that during production of the final copyedited article, errors may be discovered which could affect the content. Purpose To develop a deep learning model for the morphologic measurement of unruptured intracranial aneurysms (UIAs) based on CT angiography (CTA) data and validate its performance using a multicenter dataset. Materials and Methods In this retrospective study, patients with CTA examinations, including those with and without UIAs, in a tertiary referral hospital from February 2018 to February 2021 were included as the training dataset. Patients with UIAs who underwent CTA at multiple centers between April 2021 to December 2022 were included as the multicenter external testing set. An integrated deep-learning (IDL) model was developed for UIA detection, segmentation and morphologic measurement using an nnU-net algorithm. Model performance was evaluated using the Dice similarity coefficient (DSC) and intraclass correlation coefficient (ICC), with measurements by senior radiologists serving as the reference standard. The ability of the IDL model to improve performance of junior radiologists in measuring morphologic UIA features was assessed. Results The study included 1182 patients with UIAs and 578 controls without UIAs as the training dataset (55 years [IQR, 47-62], 1,012 [57.5%] females) and 535 patients with UIAs as the multicenter external testing set (57 years [IQR, 50-63], 353 [66.0%] females). The IDL model achieved 97% accuracy in detecting UIAs and achieved a DSC of 0.90 (95%CI, 0.88-0.92) for UIA segmentation. Model-based morphologic measurements showed good agreement with reference standard measurements (all ICCs > 0.85). Within the multicenter external testing set, the IDL model also showed agreement with reference standard measurements (all ICCs > 0.80). Junior radiologists assisted by the IDL model showed significantly improved performance in measuring UIA size (ICC improved from 0.88 [0.80-0.92] to 0.96 [0.92-0.97], P < .001). Conclusion The developed integrated deep learning model using CTA data showed good performance in UIA detection, segmentation and morphologic measurement and may be used to assist less experienced radiologists in morphologic analysis of UIAs. ©RSNA, 2024.

Predicting temporomandibular disorders in adults using interpretable machine learning methods: a model development and validation study

IntroductionTemporomandibular disorders (TMD) have a high prevalence and complex etiology. The purpose of this study was to apply a machine learning (ML) approach to identify risk factors for the occurrence of TMD in adults and to develop and validate an interpretable predictive model for the risk of TMD in adults.MethodsA total of 949 adults who underwent oral examinations were enrolled in our study. 5 different ML algorithms were used for model development and comparison, and feature selection was performed by feature importance ranking and feature decreasing methods. Several evaluation indexes, including the area under the receiver-operating-characteristic curve (AUC), were used to compare the predictive performance. The precision-recall curve (PR), calibration curve, and decision curve analysis (DCA) further assessed the accuracy and clinical utility of the model.ResultsThe performance of the random forest (RF) model was the best among the 5 ML models. An interpretable RF model was developed with 7 features (gender, malocclusion, unilateral chewing, chewing hard substances, grinding teeth, clenching teeth, and anxiety). The AUCs of the final model on the training set, internal validation set, and external test set were 0.892, 0.854, and 0.857, respectively. Calibration and DCA curves showed high accuracy and clinical applicability of the model.DiscussionAn efficient and interpretable TMD risk prediction model for adults was successfully developed using the ML method. The model not only has good predictive performance, but also enhances the clinical application value of the model through the SHAP method. This model can provide clinicians with a practical and efficient TMD risk assessment tool that can help them better predict and assess TMD risk in adults, supporting more efficient disease management and targeted medical interventions.

Prediction of femoral head collapse in osteonecrosis using deep learning segmentation and radiomics texture analysis of MRI

BackgroundFemoral head collapse is a critical pathological change and is regarded as turning point in disease progression in osteonecrosis of the femoral head (ONFH). In this study, we aim to build an automatic femoral head collapse prediction pipeline for ONFH based on magnetic resonance imaging (MRI) radiomics.MethodsIn the segmentation model development dataset, T1-weighted MRI of 222 hips from two hospitals were retrospectively collected and randomly split into training (n = 190) and test (n = 32) sets. In the prognosis prediction model development dataset, 206 hips were also retrospectively collected from two hospitals and divided into training set (n = 155) and external test set (n = 51) according to data source. A deep learning model for automatic lesion segmentation was trained with nnU-Net, from which three-dimensional regions of interest were segmented and a total of 107 radiomics features were extracted. After intra-class correlation coefficients screening, feature correlation coefficient screening and Least Absolute Shrinkage and Selection Operator regression feature selection, a machine learning model for ONFH prognosis prediction was trained with Logistic Regression (LR) and Light Gradient Boosting Machine (LightGBM) algorithm.ResultsThe segmentation model achieved an average dice similarity coefficient of 0.848 and an average 95% Hausdorff distance of 3.794 in the test set, compared to the manual segmentation results. After feature selection, nine radiomics features were included in the prognosis prediction model. External test showed that the LightGBM model exhibited acceptable predictive performance. The area under the curve (AUC) of the prediction model was 0.851 (95% CI: 0.7268–0.9752), with an accuracy of 0.765, sensitivity of 0.833, and specificity of 0.727. Decision curve analysis showed that the LightGBM model exhibited favorable clinical utility.ConclusionThis study presents an automated pipeline for predicting femoral head collapse in ONFH with acceptable performance. Further research is necessary to determine the clinical applicability of this radiomics-based approach and to assess its potential to assist in treatment decision-making for ONFH.

An interpretable deep learning model for detecting BRCA pathogenic variants of breast cancer from hematoxylin and eosin-stained pathological images.

Determining the status of breast cancer susceptibility genes (BRCA) is crucial for guiding breast cancer treatment. Nevertheless, the need for BRCA genetic testing among breast cancer patients remains unmet due to high costs and limited resources. This study aimed to develop a Bi-directional Self-Attention Multiple Instance Learning (BiAMIL) algorithm to detect BRCA status from hematoxylin and eosin (H&E) pathological images. A total of 319 histopathological slides from 254 breast cancer patients were included, comprising two dependent cohorts. Following image pre-processing, 633,484 tumor tiles from the training dataset were employed to train the self-developed deep-learning model. The performance of the network was evaluated in the internal and external test sets. BiAMIL achieved AUC values of 0.819 (95% CI [0.673-0.965]) in the internal test set, and 0.817 (95% CI [0.712-0.923]) in the external test set. To explore the relationship between BRCA status and interpretable morphological features in pathological images, we utilized Class Activation Mapping (CAM) technique and cluster analysis to investigate the connections between BRCA gene mutation status and tissue and cell features. Significantly, we observed that tumor-infiltrating lymphocytes and the morphological characteristics of tumor cells appeared to be potential features associated with BRCA status. An interpretable deep neural network model based on the attention mechanism was developed to predict the BRCA status in breast cancer. Keywords: Breast cancer, BRCA, deep learning, self-attention, interpretability.

Automatic identification of coronary stent in coronary calcium scoring CT using deep learning

Coronary artery calcium (CAC) scoring CT is a useful tool for screening coronary artery disease and for cardiovascular risk stratification. However, its efficacy in patients with coronary stents, who had pre-existing coronary artery disease, remains uncertain. Historically, CAC CT scans of these patients have been manually excluded from the CAC scoring process, even though most of the CAC scoring process is now fully automated. Therefore, we hypothesized that automating the filtering of patients with coronary stents using artificial intelligence could streamline the entire CAC workflow, eliminating the need for manual intervention. Consequently, we aimed to develop and evaluate a deep learning-based coronary stent filtering algorithm (StentFilter) in CAC scoring CT scans using a multicenter CAC dataset. We developed StentFilter comprising two main processes: stent identification and false-positive reduction. Development utilized 108 non-enhanced echocardiography-gated CAC scans (including 74 with manually labeled stents), and for false positive reduction, 2063 CAC scans with significant coronary calcium (average Agatston score: 523.8) but no stents were utilized. StentFilter’s performance was evaluated on two independent internal test sets (Asan cohort- and 2; n = 355 and 396; one without coronary stents) and two external test sets from different institutions (n = 105 and 62), each with manually labeled stents. We calculated the per-patient sensitivity, specificity, and false-positive rate of StentFilter. StentFilter demonstrated a high overall per-patient sensitivity of 98.8% (511/517 cases with stents) and a false-positive rate of 0.022 (20/918). Notably, the false-positive ratio was significantly lower in the dataset containing stents (Asan cohort-1; 0.008 [3/355]) compared with the dataset without stents (Asan cohort-2; 0.043 [17/396], p = 0.008). All false-positive identifications were attributed to dense coronary calcifications, with no false positives identified in extracoronary locations. The automated StentFilter accurately distinguished coronary stents from pre-existing coronary calcifications. This approach holds potential as a filter within a fully automated CAC scoring workflow, streamlining the process efficiently.

Construction and validation of a line chart for gestational diabetes mellitus based on clinical indicators

BackgroundGestational diabetes mellitus (GDM) is a common complication of mid–to-late pregnancy. Here, we constructed a predictive model for GDM based on a combination of clinical characteristics and relevant serum markers.MethodsData from full-term singleton vaginal deliveries from January 2022 to January 2023 were retrospectively collected from the obstetrics department. The data collected were segregated and assigned to training, validation, and external test sets. Maternal demographic characteristics, living and working habits, and haematological indicators, such as liver function and lipids were collected using a questionnaire designed for the study. The “rms” package in R was used to explore GDM-associated factors through stepwise regression at P < 0.05. A predictive model was developed based on the results of multifactorial logistic regression analysis. We then evaluated the differentiation of the column-line graphical model and performed internal and external validation. To assess the accuracy of the bar graphical model, we plotted calibration and decision curves.ResultsData from 265 pregnant women were included in the training and internal validation sets, and data from 113 pregnant women were included in the external validation set. The logistic regression algorithm screened 8 indicators as predictors. A prediction model was constructed with ALT, TBA, TC, and TG levels while considering whether GDM affects appetite, the husband– wife relationship, family history, and parental relationships as predictors. The Hosmer–Lemeshow goodness-of-fit test revealed that the chi-square values for the modelling, internal validation, and external validation groups (χ2 = 5.964, 3.249, and 12.182, respectively) were all P > 0.05. The ROC curve AUCs for the three groups were 0.93 (95% CI: 0.89–0.97), 0.72 (95% CI: 0.62–0.81), and 0.68 (95% CI: 0.53–0.83), respectively.ConclusionIn this study, a GDM prediction model was constructed to achieve high performance in GDM risk prediction based on routine obstetric tests and information.

Radiomics-Based Machine Learning with Natural Gradient Boosting for Continuous Survival Prediction in Glioblastoma

(1) Background: Glioblastoma (GBM) is the most common primary malignant brain tumor in adults, with an aggressive disease course that requires accurate prognosis for individualized treatment planning. This study aims to develop and evaluate a radiomics-based machine learning (ML) model to estimate overall survival (OS) for patients with GBM using pre-treatment multi-parametric magnetic resonance imaging (MRI). (2) Methods: The MRI data of 865 patients with GBM were assessed, comprising 499 patients from the UPENN-GBM dataset and 366 patients from the UCSF-PDGM dataset. A total of 14,598 radiomic features were extracted from T1, T1 with contrast, T2, and FLAIR MRI sequences using PyRadiomics. The UPENN-GBM dataset was used for model development (70%) and internal validation (30%), while the UCSF-PDGM dataset served as an external test set. The NGBoost Survival model was developed to generate continuous probability estimates as well as predictions for 6-, 12-, 18-, and 24-month OS. (3) Results: The NGBoost Survival model successfully predicted survival, achieving a C-index of 0.801 on internal validation and 0.725 on external validation. For 6-month OS, the model attained an AUROC of 0.791 (95% CI: 0.742–0.832) and 0.708 (95% CI: 0.654–0.748) for internal and external validation, respectively. (4) Conclusions: The radiomics-based ML model demonstrates potential to improve the prediction of OS for patients with GBM.

Integrating VAI-Assisted Quantified CXRs and Multimodal Data to Assess the Risk of Mortality.

To address the unmet need for a widely available examination for mortality prediction, this study developed a foundation visual artificial intelligence (VAI) to enhance mortality risk stratification using chest X-rays (CXRs). The VAI employed deep learning to extract CXR features and a Cox proportional hazard model to generate a hazard score ("CXR-risk"). We retrospectively collected CXRs from patients visited outpatient department and physical examination center. Subsequently, we reviewed mortality and morbidity outcomes from electronic medical records. The dataset consisted of 41,945, 10,492, 31,707, and 4441 patients in the training, validation, internal test, and external test sets, respectively. During the median follow-up of 3.2 (IQR, 1.2-6.1) years of both internal and external test sets, the "CXR-risk" demonstrated C-indexes of 0.859 (95% confidence interval (CI), 0.851-0.867) and 0.870 (95% CI, 0.844-0.896), respectively. Patients with high "CXR-risk," above 85th percentile, had a significantly higher risk of mortality than those with low risk, below 50th percentile. The addition of clinical and laboratory data and radiographic report further improved the predictive accuracy, resulting in C-indexes of 0.888 and 0.900. The VAI can provide accurate predictions of mortality and morbidity outcomes using just a single CXR, and it can complement other risk prediction indicators to assist physicians in assessing patient risk more effectively.

Risk stratification of thymic epithelial tumors based on peritumor CT radiomics and semantic features.

To develop and validate nomograms combining radiomics and semantic features to identify the invasiveness and histopathological risk stratification of thymic epithelial tumors (TET) using contrast-enhanced CT. This retrospective multi-center study included 224 consecutive cases. For each case, 6764 intratumor and peritumor radiomics features and 31 semantic features were collected. Multi-feature selections and decision tree models were performed on radiomics features and semantic features separately to select the most important features for Masaoka-Koga staging and WHO classification. The selected features were then combined to create nomograms for the two systems. The performance of the radiomics model, semantic model, and combined model was evaluated using the area under the receiver operating characteristic curves (AUCs). One hundred eighty-seven cases (56.5 years ± 12.3, 101 men) were included, with 62 cases as the external test set. For Masaoka-Koga staging, the combined model, which incorporated five peritumor radiomics features and four semantic features, showed an AUC of 0.958 (95% CI: 0.912-1.000) in distinguishing between early-stage (stage I/II) and advanced-stage (III/IV) TET in the external test set. For WHO classification, the combined model incorporating five peritumor radiomics features and two semantic features showed an AUC of 0.857 (0.760-0.955) in differentiating low-risk (type A/AB/B1) and high-risk (B2/B3/C) TET. The combined models showed the most effective predictive performance, while the semantic models exhibited comparable performance to the radiomics models in both systems (p > 0.05). The nomograms combining peritumor radiomics features and semantic features could help in increasing the accuracy of grading invasiveness and risk stratification of TET. Peripheral invasion and histopathological type are major determinants of treatment and prognosis of TET. The integration of peritumoral radiomics features and semantic features into nomograms may enhance the accuracy of grading invasiveness and risk stratification of TET. Peritumor region of TET may suggest histopathological and invasive risk. Peritumor radiomic and semantic features allow classification by Masaoka-Koga staging (AUC: 0.958). Peritumor radiomic and semantic features enable the classification of histopathological risk (AUC: 0.857).

Personalized three-year survival prediction and prognosis forecast by interpretable machine learning for pancreatic cancer patients: a population-based study and an external validation.

The overall survival of patients with pancreatic cancer is extremely low. We aimed to establish machine learning (ML) based model to accurately predict three-year survival and prognosis of pancreatic cancer patients. We analyzed pancreatic cancer patients from the Surveillance, Epidemiology, and End Results (SEER) database between 2000 and 2021. Univariate and multivariate logistic analysis were employed to select variables. Recursive Feature Elimination (RFE) method based on 6 ML algorithms was utilized in feature selection. To construct predictive model, 13 ML algorithms were evaluated by area under the curve (AUC), area under precision-recall curve (PRAUC), accuracy, sensitivity, specificity, precision, cross-entropy, Brier scores and Balanced Accuracy (bacc) and F Beta Score (fbeta). An optimal ML model was constructed to predict three-year survival, and the predictive results were explained by SHapley Additive exPlanations (SHAP) framework. Meanwhile, 101 ML algorithm combinations were developed to select the best model with highest C-index to predict prognosis of pancreatic cancer patients. A total of 20,064 pancreatic cancer patients from SEER database was consecutively enrolled. We utilized eight clinical variables to establish prediction model for three-year survival. CatBoost model was selected as the best prediction model, and AUC was 0.932 [0.924, 0.939], 0.899 [0.873, 0.934] and 0.826 [0.735, 0.919] in training, internal test and external test sets, with 0.839 [0.831, 0.847] accuracy, 0.872 [0.858, 0.887] sensitivity, 0.803 [0.784, 0.825] specificity and 0.832 [0.821, 0.853] precision. Surgery type had the greatest effects on three-year survival according to SHAP results. For prognosis prediction, "RSF+GBM" algorithm was the best prognostic model with C-index of 0.774, 0.722 and 0.674 in training, internal test and external test sets. Our ML models demonstrate excellent accuracy and reliability, offering more precise personalized prognostic prediction to pancreatic cancer patients.

A multi-model approach integrating whole-slide imaging and clinicopathologic features to predict breast cancer recurrence risk.

Breast cancer is the most common malignancy affecting women worldwide and is notable for its morphologic and biologic diversity, with varying risks of recurrence following treatment. The Oncotype DX Breast Recurrence Score test is an important predictive and prognostic genomic assay for estrogen receptor positive/HER2 negative breast cancer that guides therapeutic strategies; however, such tests can be expensive, delay care, and are not widely available. The aim of this study was to develop a multi-model approach integrating the analysis of whole-slide images and clinicopathologic data to predict their associated breast cancer recurrence risks and categorize these patients into two risk groups according to the predicted score: low-risk and high-risk. The proposed novel methodology uses convolutional neural networks for feature extraction and vision transformers for contextual aggregation, complemented by a logistic regression model that analyzes clinicopathologic data for classification into two risk categories. This method was trained and tested on 956 hematoxylin and eosin-stained whole-slide images of 950 ER+/HER2- breast cancer patients with corresponding clinicopathological features that had prior Oncotype DX testing. The model's performance was evaluated using an internal test set of 192 patients from Dartmouth Health and an external test set of 405 patients from the University of Chicago. The multi-model approach achieved an AUC of 0.91 (95% CI: 0.87-0.95) on the internal set and an AUC of 0.84 (95% CI: 0.78-0.89) on the external cohort for predicting low- and high-breast cancer recurrence risk categories based on the Oncotype DX recurrence score. With further validation, the proposed methodology could provide an alternative to assist clinicians in personalizing treatment for breast cancer patients and potentially improving their outcomes.

Multiparametric MRI Radiomics With Machine Learning for Differentiating HER2-Zero, -Low, and -Positive Breast Cancer: Model Development, Testing, and Interpretability Analysis.

BACKGROUND: MRI radiomics has been explored for three-tiered classification of breast cancer HER2 expression (i.e., HER2-zero, HER2-low, or HER2-positive), although understanding of how such models reach their predictions is lacking. OBJECTIVE: To develop and test multiparametric MRI radiomics machine-learning models for differentiating three-tiered HER2 expression levels in patients with breast cancer, and to explain the contributions of model features through local and global interpretations using SHapley Additive exPlanation (SHAP) analysis. METHODS: This retrospective study included 737 patients (mean age, 54.1±10.6 years) with breast cancer from two centers (center 1: n=578; center 2: n=159), who underwent breast MRI and had HER2 expression determined after excisional biopsy. Analysis entailed two tasks: differentiating HER2-negative (i.e., HER2-zero or HER2-low) from HER2-positive tumors (task 1), and differentiating HER2-zero from HER2-low tumors (task 2). For each task, patients from center 1 were randomly assigned in 7:3 ratio to training (task 1: n=405; task 2: n=284) or internal test (task 1: n=173; task 2: n=122) sets; those from center 2 formed an external test set (task 1: n=159; task 2: n=105). Radiomics features were extracted from early-phase dynamic contrast-enhanced images (DCE), T2-weighted images (T2WI), and DWI. For each task, a support vector machine (SVM) was used for feature selection; a multiparametric radiomics score (radscore) was computed using feature weights from SVM correlation coefficients; conventional MRI and combined models were constructed; and model performances were evaluated. SHAP analysis was used to provide local and global interpretations for model outputs. RESULTS: In the external test set, for task 1, AUCs for the conventional MRI model, radscore, and combined model were 0.624, 0.757, and 0.762, respectively; for task 2, AUC for radscore was 0.754, and no conventional MRI model or combined model could be constructed. SHAP analysis identified early-phase DCE features as having the strongest influence for both tasks; T2WI features also had a prominent role for task 2. CONCLUSION: The findings indicate suboptimal performance of MRI radiomics models for noninvasive characterization of HER2 expression. CLINICAL IMPACT: The study provides an example of the use of SHAP interpretation analysis to better understand predictions of imaging-based machine learning models.

A deep learning model to enhance the classification of primary bone tumors based on incomplete multimodal images in X-ray, CT, and MRI

BackgroundAccurately classifying primary bone tumors is crucial for guiding therapeutic decisions. The National Comprehensive Cancer Network guidelines recommend multimodal images to provide different perspectives for the comprehensive evaluation of primary bone tumors. However, in clinical practice, most patients’ medical multimodal images are often incomplete. This study aimed to build a deep learning model using patients’ incomplete multimodal images from X-ray, CT, and MRI alongside clinical characteristics to classify primary bone tumors as benign, intermediate, or malignant.MethodsIn this retrospective study, a total of 1305 patients with histopathologically confirmed primary bone tumors (internal dataset, n = 1043; external dataset, n = 262) were included from two centers between January 2010 and December 2022. We proposed a Primary Bone Tumor Classification Transformer Network (PBTC-TransNet) fusion model to classify primary bone tumors. Areas under the receiver operating characteristic curve (AUC), accuracy, sensitivity, and specificity were calculated to evaluate the model’s classification performance.ResultsThe PBTC-TransNet fusion model achieved satisfactory micro-average AUCs of 0.847 (95% CI: 0.832, 0.862) and 0.782 (95% CI: 0.749, 0.817) on the internal and external test sets. For the classification of benign, intermediate, and malignant primary bone tumors, the model respectively achieved AUCs of 0.827/0.727, 0.740/0.662, and 0.815/0.745 on the internal/external test sets. Furthermore, across all patient subgroups stratified by the distribution of imaging modalities, the PBTC-TransNet fusion model gained micro-average AUCs ranging from 0.700 to 0.909 and 0.640 to 0.847 on the internal and external test sets, respectively. The model showed the highest micro-average AUC of 0.909, accuracy of 84.3%, micro-average sensitivity of 84.3%, and micro-average specificity of 92.1% in those with only X-rays on the internal test set. On the external test set, the PBTC-TransNet fusion model gained the highest micro-average AUC of 0.847 for patients with X-ray + CT.ConclusionsWe successfully developed and externally validated the transformer-based PBTC-Transnet fusion model for the effective classification of primary bone tumors. This model, rooted in incomplete multimodal images and clinical characteristics, effectively mirrors real-life clinical scenarios, thus enhancing its strong clinical practicability.

Pathologist-level diagnosis of ulcerative colitis inflammatory activity level using an automated histological grading method

Background and AimsInflammatory bowel disease (IBD) is a global disease that is evolving with increasing incidence. However, there are few works on computationally assisted diagnosis of IBD based on pathological images. Therefore, based on the UK and Chinese IBD diagnostic guidelines, our study established an artificial intelligence-assisted diagnostic system for histologic grading of inflammatory activity in ulcerative colitis (UC). MethodsWe proposed an efficient deep-learning (DL) method for grading inflammatory activity in whole-slide images (WSIs) of UC pathology. Our model was constructed using 603 UC WSIs from Nanjing Drum Tower Hospital for model train set and internal test set. We collected 212 UC WSIs from Zhujiang Hospital as an external test set. Initially, the pre-trained ResNet50 model on the ImageNet dataset was employed to extract image patch features from UC patients. Subsequently, a multi-instance learning (MIL) approach with embedded self-attention was utilized to aggregate tissue image patch features, representing the entire WSI. Finally, the model was trained based on the aggregated features and WSI annotations provided by senior gastrointestinal pathologists to predict the level of inflammatory activity in UC WSIs. ResultsIn the task of distinguishing the presence or absence of inflammatory activity, the Area Under Curve (AUC) value in the internal test set is 0.863 (95% confidence interval [CI] 0.829, 0.898), with a sensitivity of 0.913 (95% [CI] 0.866, 0.961), and specificity of 0.816 (95% [CI] 0.771, 0.861). The AUC in the external test set is 0.947 (95% confidence interval [CI] 0.939, 0.955), with a sensitivity of 0.889 (905% [CI] 0.837, 0.940), and specificity of 0.858 (95% [CI] 0.777, 0.939). For distinguishing different levels of inflammatory activity in UC, the average Macro-AUC in the internal test set and the external test set are 0.827 (95% [CI] 0.803, 0.850) and 0.908 (95% [CI] 0.882, 0.935). the average Micro-AUC in the internal test set and the external test set are 0.816 (95% [CI] 0.792, 0.840) and 0.898 (95% [CI] 0.869, 0.926). ConclusionsComparative analysis with diagnoses made by pathologists at different expertise levels revealed that the algorithm reached a proficiency comparable to the pathologist with 5 years of experience. Furthermore, our algorithm performed superior to other MIL algorithms.

Prediction of benign and malignant ground glass pulmonary nodules based on multi-feature fusion of attention mechanism.

The purpose of this study was to develop and validate a new feature fusion algorithm to improve the classification performance of benign and malignant ground-glass nodules (GGNs) based on deep learning. We retrospectively collected 385 cases of GGNs confirmed by surgical pathology from three hospitals. We utilized 239 GGNs from Hospital 1 as the training and internal validation set, and 115 and 31 GGNs from Hospital 2 and Hospital 3, respectively, as external test sets 1 and 2. Among these GGNs, 172 were benign and 203 were malignant. First, we evaluated clinical and morphological features of GGNs at baseline chest CT and simultaneously extracted whole-lung radiomics features. Then, deep convolutional neural networks (CNNs) and backpropagation neural networks (BPNNs) were applied to extract deep features from whole-lung CT images, clinical, morphological features, and whole-lung radiomics features separately. Finally, we integrated these four types of deep features using an attention mechanism. Multiple metrics were employed to evaluate the predictive performance of the model. The deep learning model integrating clinical, morphological, radiomics and whole lung CT image features with attention mechanism (CMRI-AM) achieved the best performance, with area under the curve (AUC) values of 0.941 (95% CI: 0.898-0.972), 0.861 (95% CI: 0.823-0.882), and 0.906 (95% CI: 0.878-0.932) on the internal validation set, external test set 1, and external test set 2, respectively. The AUC differences between the CMRI-AM model and other feature combination models were statistically significant in all three groups (all p<0.05). Our experimental results demonstrated that (1) applying attention mechanism to fuse whole-lung CT images, radiomics features, clinical, and morphological features is feasible, (2) clinical, morphological, and radiomics features provide supplementary information for the classification of benign and malignant GGNs based on CT images, and (3) utilizing baseline whole-lung CT features to predict the benign and malignant of GGNs is an effective method. Therefore, optimizing the fusion of baseline whole-lung CT features can effectively improve the classification performance of GGNs.

Deep learning model using planar whole-body bone scintigraphy for diagnosis of skull base invasion in patients with nasopharyngeal carcinoma

PurposeThis study assesses the reliability of deep learning models based on planar whole-body bone scintigraphy for diagnosing Skull base invasion (SBI) in nasopharyngeal carcinoma (NPC) patients.MethodsIn this multicenter study, a deep learning model was developed using data from one center with a 7:3 allocation to training and internal test sets, to diagnose SBI in patients newly diagnosed with NPC using planar whole-body bone scintigraphy. Patients were diagnosed based on a composite reference standard incorporating radiologic and follow-up data. Ten different convolutional neural network (CNN) models were applied to both whole-image and partial-image input modes to determine the optimal model for each analysis. Model performance was assessed using the area under the receiver operating characteristic curve (AUC), calibration, decision curve analysis (DCA), and compared with expert assessments by two nuclear medicine physicians.ResultsThe best-performing model using partial-body input achieved AUCs of 0.80 (95% CI: 0.73, 0.86) in the internal test set, 0.84 (95% CI: 0.77, 0.91) in the external cohort, and 0.78 (95% CI: 0.73, 0.83) in the treatment test cohort. Calibration curves and DCA confirmed the models’ excellent discrimination, calibration, and potential clinical utility across internal and external datasets. The AUCs of both nuclear medicine physicians were lower than those of the best-performing deep learning model in external test set (AUC: 0.75 vs. 0.77 vs. 0.84).ConclusionDeep learning models utilizing partial-body input from planar whole-body bone scintigraphy demonstrate high discriminatory power for diagnosing SBI in NPC patients, surpassing experienced nuclear medicine physicians.

TransCDR: a deep learning model for enhancing the generalizability of drug activity prediction through transfer learning and multimodal data fusion

BackgroundAccurate and robust drug response prediction is of utmost importance in precision medicine. Although many models have been developed to utilize the representations of drugs and cancer cell lines for predicting cancer drug responses (CDR), their performances can be improved by addressing issues such as insufficient data modality, suboptimal fusion algorithms, and poor generalizability for novel drugs or cell lines.ResultsWe introduce TransCDR, which uses transfer learning to learn drug representations and fuses multi-modality features of drugs and cell lines by a self-attention mechanism, to predict the IC50 values or sensitive states of drugs on cell lines. We are the first to systematically evaluate the generalization of the CDR prediction model to novel (i.e., never-before-seen) compound scaffolds and cell line clusters. TransCDR shows better generalizability than 8 state-of-the-art models. TransCDR outperforms its 5 variants that train drug encoders (i.e., RNN and AttentiveFP) from scratch under various scenarios. The most critical contributors among multiple drug notations and omics profiles are Extended Connectivity Fingerprint and genetic mutation. Additionally, the attention-based fusion module further enhances the predictive performance of TransCDR. TransCDR, trained on the GDSC dataset, demonstrates strong predictive performance on the external testing set CCLE. It is also utilized to predict missing CDRs on GDSC. Moreover, we investigate the biological mechanisms underlying drug response by classifying 7675 patients from TCGA into drug-sensitive or drug-resistant groups, followed by a Gene Set Enrichment Analysis.ConclusionsTransCDR emerges as a potent tool with significant potential in drug response prediction.

ProstateZones – Segmentations of the prostatic zones and urethra for the PROSTATEx dataset

Manual segmentations are considered the gold standard for ground truth in machine learning applications. Such tasks are tedious and time-consuming, albeit necessary to train reliable models. In this work, we present a dataset with expert segmentations of the prostatic zones and urethra for 200 randomly selected patients from the PROSTATEx dataset. Notably, independent duplicate segmentations were performed for 40 patients, providing inter-reader variability data. This results in a total of 240 segmentations. This dataset can be used to train machine learning models or serve as an external test set for evaluating models trained on private data, thereby addressing a current gap in the field. The delineated structures and terminology adhere to the latest Prostate Imaging Reporting and Data Systems v2.1 guidelines, ensuring consistency.

MRI-based radiomics for stratifying recurrence risk of early-onset rectal cancer: a multicenter study

Early-onset rectal cancer (EORC) is characterized by a unique disease process with different clinicopathological features compared with late-onset rectal cancer (LORC). Research on the risk of recurrence in EORC patients, however, is limited. We aim to develop a predictive model to accurately predict EORC recurrence risk. Rectal cancer patients who underwent radical surgery and T2-weighted imaging and diffusion-weighted imaging magnetic resonance imaging (MRI) were retrospectively enrolled from three medical institutions from November 2012 to November 2018. Differences in clinicopathological characteristics between EORC and LORC were compared. Five prediction models for disease-free survival were constructed based on clinicopathological variables and five radiomic features from pretreatment MRI of the EORC. A fixed cut-off value calculated in the training set was used to stratify EORC patients into high-risk and low-risk groups of post-operative recurrence. Model performance was evaluated by concordance index (C-index) and receiver operating characteristic curve. A total of 264 EORC patients (median age, 43 years, 163 males) and 778 LORC patients (median age, 62 years, 520 males) were enrolled. Pretreatment positive carcinoembryonic antigen [hazard ratio (HR)= 2.84, P= 0.006], pathological positive lymph node status (pN positive) [HR= 2.86, P= 0.011] and MRI-based radiomics score [HR= 2.72, P < 0.001] are independent risk factors for disease-free survival in EORC patients. The EORC-ClinPathRadiom model, constructed by integrating the clinicopathological characteristics and MRI-based radiomics features of EORC, showed C-index of 0.82, 0.82, and 0.81 in the training, internal, and external test sets, respectively. This model effectively stratified EORC patients into high risk and low risk of recurrence (HRs for the training, internal, and external test sets were 8.96, 6.81, and 7.46, respectively). The EORC-ClinPathRadiom model can effectively predict and stratify the risk of post-operative recurrence in EORC patients.

Risk-stratified management of cervical high-grade squamous intraepithelial lesion based on machine learning.

The concordance rate between conization and colposcopy-directed biopsy (CDB) proven cervical high-grade squamous intraepithelial lesion (HSIL) were 64-85%. We aimed to identify the risk factors associated with pathological upgrading or downgrading after conization in patients with cervical HSIL and to provide risk-stratified management based on a machine learning predictive model. This retrospective study included patients who visited the Obstetrics and Gynecology Hospital of Fudan University from January 1 to December 31, 2019, were diagnosed with cervical HSIL by CDB, and subsequently underwent conization. A wide variety of data were collected from the medical records, including demographic data, laboratory findings, colposcopy descriptions, and pathological results. The patients were categorized into three groups according to their postconization pathological results: low-grade squamous intraepithelial lesion (LSIL) or below (downgrading group), HSIL (HSIL group), and cervical cancer (upgrading group). Univariate and multivariate analyses were performed to identify the independent risk factors for pathological changes in patients with cervical HSIL. Machine learning prediction models were established, evaluated, and subsequently verified using external testing data. In total, 1585 patients were included, of whom 65 (4.1%) were upgraded to cervical cancer after conization, 1147 (72.4%) remained having HSIL, and 373 (23.5%) were downgraded to LSIL or below. Multivariate analysis showed a 2% decrease in the incidence of pathological downgrade for each additional year of age and a 1% increase in lesion size. Patients with cytology > LSIL (odds ratio [OR] = 0.33; 95% confidence interval [CI], 0.21-0.52), human papillomavirus (HPV) infection (OR = 0.33; 95% CI, 0.14-0.81), HPV 33 infection (OR = 0.37; 95% CI, 0.18-0.78), coarse punctate vessels on colposcopy examination (OR = 0.14; 95% CI, 0.06-0.32), HSIL lesions in the endocervical canal (OR = 0.48; 95% CI, 0.30-0.76), and HSIL impression (OR = 0.02; 95% CI, 0.01-0.03) were less likely to experience pathological downgrading after conization than their counterparts. The independent risk factors for pathological upgrading to cervical cancer after conization included the following: age (OR = 1.08; 95% CI, 1.04-1.12), HPV 16 infection (OR = 4.07; 95% CI, 1.70-9.78), the presence of coarse punctate vessels during colposcopy examination (OR = 2.21; 95% CI, 1.08-4.50), atypical vessels (OR = 6.87; 95% CI, 2.81-16.83), and HSIL lesions in the endocervical canal (OR = 2.91; 95% CI, 1.46-5.77). Among the six machine learning prediction models, the back propagation (BP) neural network model demonstrated the highest and most uniform predictive performance in the downgrading, HSIL, and upgrading groups, with areas under the curve (AUCs) of 0.90, 0.84, and 0.69; sensitivities of 0.74, 0.84, and 0.42; specificities of 0.90, 0.71, and 0.95; and accuracies of 0.74, 0.84, and 0.95, respectively. In the external testing set, the BP neural network model showed a higher predictive performance than the logistic regression model, with an overall AUC of 0.91. Therefore, a web-based prediction tool was developed in this study. BP neural network prediction model has excellent predictive performance and can be used for the risk stratification of patients with CDB-diagnosed HSIL.