External Test Data Research Articles

Automated measurement of cardiothoracic ratio based on semantic segmentation integration model using deep learning.

The objective of this study is to investigate the efficacy of the semantic segmentation model in predicting cardiothoracic ratio (CTR) and heart enlargement and compare its consistency with the reference standard. A total of 650 consecutive chest radiographs from our center and 756 public datasets were retrospectively included to develop a segmentation model. Three semantic segmentation models were used to segment the heart and lungs. A soft voting integration method was used to improve the segmentation accuracy and measure CTR automatically. Bland-Altman and Pearson's correlation analyses were used to compare the consistency and correlation between CTR automated measurements and reference standards. CTR automated measurements were compared with reference standard using the Wilcoxon signed-rank test. The diagnostic efficacy of the model for heart enlargement was evaluated using the AUC. The soft voting integration model was strongly correlated (r = 0.98, P < 0.001) and consistent (average standard deviation of 0.0048cm/s) with the reference standard. No statistical difference between CTR automated measurement and reference standard in healthy subjects, pneumothorax, pleural effusion, and lung mass patients (P > 0.05). In the external test data, the accuracy, sensitivity, specificity, and AUC in determining heart enlargement were 96.0%, 79.5%, 99.1%, and 0.988, respectively. The deep learning method was calculated faster per chest radiograph than the average time manually calculated by the radiologist (about 2s vs 25.75 ± 4.35s, respectively, P < 0.001). This study provides a semantic segmentation integration model of chest radiographs to measure CTR and determine heart enlargement with chest structure changes due to different chest diseases effectively, faster, and accurately. The development of the automated segmentation integration model is helpful in improving the consistency of CTR measurement, reducing the workload of radiologists, and improving their work efficiency.

Artificial intelligence-based personalized survival prediction using clinical and radiomics features in patients with advanced non-small cell lung cancer

BackgroundMultiple first-line treatment options have been developed for advanced non-small cell lung cancer (NSCLC) in each subgroup determined by predictive biomarkers, specifically driver oncogene and programmed cell death ligand-1 (PD-L1) status. However, the methodology for optimal treatment selection in individual patients is not established. This study aimed to develop artificial intelligence (AI)-based personalized survival prediction model according to treatment selection.MethodsThe prediction model was built based on random survival forest (RSF) algorithm using patient characteristics, anticancer treatment histories, and radiomics features of the primary tumor. The predictive accuracy was validated with external test data and compared with that of cox proportional hazard (CPH) model.ResultsA total of 459 patients (training, n = 299; test, n = 160) with advanced NSCLC were enrolled. The algorithm identified following features as significant factors associated with survival: age, sex, performance status, Brinkman index, comorbidity of chronic obstructive pulmonary disease, histology, stage, driver oncogene status, tumor PD-L1 expression, administered anticancer agent, six markers of blood test (sodium, lactate dehydrogenase, etc.), and three radiomics features associated with tumor texture, volume, and shape. The C-index of RSF model for test data was 0.841, which was higher than that of CPH model (0.775, P < 0.001). Furthermore, the RSF model enabled to identify poor survivor treated with pembrolizumab because of tumor PD-L1 high expression and those treated with driver oncogene targeted therapy according to driver oncogene status.ConclusionsThe proposed AI-based algorithm accurately predicted the survival of each patient with advanced NSCLC. The AI-based methodology will contribute to personalized medicine.Trial registrationThe trial design was retrospectively registered study performed in accordance with the Declaration of Helsinki and approved by the Institutional Review Board of Nagoya University Graduate School of Medicine (approval: 2020 − 0287).

Comparative efficacy of anteroposterior and lateral X-ray based deep learning in the detection of osteoporotic vertebral compression fracture

Magnetic resonance imaging remains the gold standard for diagnosing osteoporotic vertebral compression fractures (OVCF), but the use of X-ray imaging, particularly anteroposterior (AP) and lateral views, is prevalent due to its accessibility and cost-effectiveness. We aim to assess whether the performance of AP images-based deep learning is comparable compared to those using lateral images. This retrospective study analyzed X-ray images from two tertiary teaching hospitals, involving 1,507 patients for the training and internal test, and 104 patients for the external test. The EfficientNet-B5-based algorithms were employed to classify OVCF and non-OVCF group. The model was trained with a 1:1 balanced dataset and validated through 5-fold cross validation. Performance outcomes were compared with the area under receiver operating characteristic (AUROC) curve. Out of a total of 1,507 patients, 799 were included in the training dataset and 708 were included in the internal test dataset. The training and internal test datasets were matched 1:1 as OVCF and non-OVCF patients. The DL model showed comparable classifying performance with internal test data (N = 708, AUROC for AP, 0.915; AUROC for lateral, 0.953) and external test data (N = 104, AUROC for AP, 0.982; AUROC for lateral, 0979), respectively. The other performances including F1 score and accuracy were also comparable. Especially, The AUROC of AP and lateral x-ray image-based DL was not significantly different (p for DeLong test = 0.604). The EfficientNet-B5 algorithms using AP X-ray images shows comparable efficacy for classifying OVCF and non-OVCF compared to lateral images.

Using Machine Learning Models to Predict Pathologic Complete Response to Neoadjuvant Chemotherapy in Breast Cancer.

Neoadjuvant chemotherapy (NAC) is increasingly used in breast cancer. Predictive modeling is useful in predicting pathologic complete response (pCR) to NAC. We test machine learning (ML) models to predict pCR in breast cancer and explore methods of handling missing data. Four hundred and ninety-nine patients with breast cancer treated with NAC in two centers in Singapore (National Cancer Centre Singapore [NCCS] and KK Hospital) between January 2014 and December 2017 were included. Eleven clinical features were used to train five different ML models. Listwise deletion and imputation were evaluated on handling missing data. Model performance was evaluated by AUC and calibration (Brier score). Feature importance from the best performing model in the external testing data set was calculated using Shapley additive explanations. Seventy-two (24.6%), 18 (24.7%), and 31 (24.8%) patients attained pCR in NCCS training, NCCS testing, and KK Women's and Children's Hospital (KKH) testing data sets, respectively. The random forest (RF) base and imputed models have the highest AUCs in the KKH cohort of 0.794 (95% CI, 0.709 to 0.873) and 0.795 (95% CI, 0.706 to 0.871), respectively, and were the best calibrated with the lowest Brier score. No statistically significant difference was noted between AUCs of the base and imputed models in all data sets. The imputed model had a larger positive predictive value (PPV; 98.2% v 95.1%) and negative predictive value (NPV; 96.7% v 90.0%) than the base model in the KKH data set. Estrogen receptor intensity, human epidermal growth factor 2 intensity, and age at diagnosis were the three most important predictors. ML, particularly RF, demonstrates reasonable accuracy in pCR prediction after NAC. Imputing missing fields in the data can improve the PPV and NPV of the pCR prediction model.

Development and Validation of a Computed Tomography-Based Model for Noninvasive Prediction of the T Stage in Gastric Cancer: Multicenter Retrospective Study.

As part of the TNM (tumor-node-metastasis) staging system, T staging based on tumor depth is crucial for developing treatment plans. Previous studies have constructed a deep learning model based on computed tomographic (CT) radiomic signatures to predict the numberoflymphnodemetastases and survival in patients with resected gastric cancer (GC). However, few studies have reported the combination of deep learning and radiomics in predicting T staging in GC. This study aimed to develop a CT-based model for automatic prediction of the T stage of GC via radiomics and deep learning. A total of 771 GC patients from 3 centers were retrospectively enrolled and divided into training, validation, and testing cohorts. Patients with GC were classified into mild (stage T1 and T2), moderate (stage T3), and severe (stage T4) groups. Three predictive models based on the labeled CT images were constructed using the radiomics features (radiomics model), deep features (deep learning model), and a combination of both (hybrid model). The overall classification accuracy of the radiomics model was 64.3% in the internal testing data set. The deep learning model and hybrid model showed better performance than the radiomics model, with overall classification accuracies of 75.7% (P=.04) and 81.4% (P=.001), respectively. On the subtasks of binary classification of tumor severity, the areas under the curve of the radiomics, deep learning, and hybrid models were 0.875, 0.866, and 0.886 in the internal testing data set and 0.820, 0.818, and 0.972 in the external testing data set, respectively, for differentiating mild (stage T1~T2) from nonmild (stage T3~T4) patients, and were 0.815, 0.892, and 0.894 in the internal testing data set and 0.685, 0.808, and 0.897 in the external testing data set, respectively, for differentiating nonsevere (stage T1~T3) from severe (stage T4) patients. The hybrid model integrating radiomics features and deep features showed favorable performance in diagnosing the pathological stage of GC.

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.

CPred: Charge State Prediction for Modified and Unmodified Peptides in Electrospray Ionization.

The mass-to-charge ratio serves as a critical parameter in peptide identification via mass spectrometry, enabling the precise determination of peptide masses and facilitating their differentiation based on unique charge characteristics, especially when peptides are ionized by tools like electrospray ionization, which produces multiply charged ions. We developed a neural network called CPred, which can accurately predict the charge state distribution from +1 to +7 for the modified and unmodified peptides. CPred was trained on the large-scale synthetic training data, consisting of tryptic and non-tryptic peptides, and various fragmentation methods. The model was further evaluated on independent, external test data sets. Results were evaluated through the Pearson correlation coefficient and showed high correlations of up to 0.9997117 between the predicted and acquired charge state distributions. The effect of specifying modifications in the neural network and feature importance was further investigated, revealing the value of modifications and vital peptide properties in holding on to protons. CPreds' accurate predictions of the charge state distribution can play an essential role in boosting confidence in peptide identifications during rescoring as a novel feature.

Renal tumor segmentation, visualization, and segmentation confidence using ensembles of neural networks in patients undergoing surgical resection.

To develop an automatic segmentation model for solid renal tumors on contrast-enhanced CTs and to visualize segmentation with associated confidence to promote clinical applicability. The training dataset included solid renal tumor patients from two tertiary centers undergoing surgical resection and receiving CT in the corticomedullary or nephrogenic contrast media (CM) phase. Manual tumor segmentation was performed on all axial CT slices serving as reference standard for automatic segmentations. Independent testing was performed on the publicly available KiTS 2019 dataset. Ensembles of neural networks (ENN, DeepLabV3) were used for automatic renal tumor segmentation, and their performance was quantified with DICE score. ENN average foreground entropy measured segmentation confidence (binary: successful segmentation with DICE score > 0.8 versus inadequate segmentation ≤ 0.8). N = 639/n = 210 patients were included in the training and independent test dataset. Datasets were comparable regarding age and sex (p > 0.05), while renal tumors in the training dataset were larger and more frequently benign (p < 0.01). In the internal test dataset, the ENN model yielded a median DICE score = 0.84 (IQR: 0.62-0.97, corticomedullary) and 0.86 (IQR: 0.77-0.96, nephrogenic CM phase), and the segmentation confidence an AUC = 0.89 (sensitivity = 0.86; specificity = 0.77). In the independent test dataset, the ENN model achieved a median DICE score = 0.84 (IQR: 0.71-0.97, corticomedullary CM phase); and segmentation confidence an accuracy = 0.84 (sensitivity = 0.86 and specificity = 0.81). ENN segmentations were visualized with color-coded voxelwise tumor probabilities and thresholds superimposed on clinical CT images. ENN-based renal tumor segmentation robustly performs in external test data and might aid in renal tumor classification and treatment planning. Ensembles of neural networks (ENN) models could automatically segment renal tumors on routine CTs, enabling and standardizing downstream image analyses and treatment planning. Providing confidence measures and segmentation overlays on images can lower the threshold for clinical ENN implementation. Ensembles of neural networks (ENN) segmentation is visualized by color-coded voxelwise tumor probabilities and thresholds. ENN provided a high segmentation accuracy in internal testing and in an independent external test dataset. ENN models provide measures of segmentation confidence which can robustly discriminate between successful and inadequate segmentations.

On the Search for Data-Driven and Reproducible Schizophrenia Subtypes Using Resting State fMRI Data From Multiple Sites.

For decades, fMRI data have been used to search for biomarkers for patients with schizophrenia. Still, firm conclusions are yet to be made, which is often attributed to the high internal heterogeneity of the disorder. A promising way to disentangle the heterogeneity is to search for subgroups of patients with more homogeneous biological profiles. We applied an unsupervised multiple co-clustering (MCC) method to identify subtypes using functional connectivity data from a multisite resting-state data set. We merged data from two publicly available databases and split the data into a discovery data set (143 patients and 143 healthy controls (HC)) and an external test data set (63 patients and 63HC) from independent sites. On the discovery data, we investigated the stability of the clustering toward data splits and initializations. Subsequently we searched for cluster solutions, also called "views," with a significant diagnosis association and evaluated these based on their subject and feature cluster separability, and correlation to clinical manifestations as measured with the positive and negative syndrome scale (PANSS). Finally, we validated our findings by testing the diagnosis association on the external test data. A major finding of our study was that the stability of the clustering was highly dependent on variations in the data set, and even across initializations, we found only a moderate subject clustering stability. Nevertheless, we still discovered one view with a significant diagnosis association. This view reproducibly showed an overrepresentation of schizophrenia patients in three subject clusters, and one feature cluster showed a continuous trend, ranging from positive to negative connectivity values, when sorted according to the proportions of patients with schizophrenia. When investigating all patients, none of the feature clusters in the view were associated with severity of positive, negative, and generalized symptoms, indicating that the cluster solutions reflect other disease related mechanisms.

Deep Learning Models for Abdominal CT Organ Segmentation in Children: Development and Validation in Internal and Heterogeneous Public Datasets.

BACKGROUND. Deep learning abdominal organ segmentation algorithms have shown excellent results in adults; validation in children is sparse. OBJECTIVE. The purpose of this article is to develop and validate deep learning models for liver, spleen, and pancreas segmentation on pediatric CT examinations. METHODS. This retrospective study developed and validated deep learning models for liver, spleen, and pancreas segmentation using 1731 CT examinations (1504 training, 221 testing), derived from three internal institutional pediatric (age ≤ 18 years) datasets (n = 483) and three public datasets comprising pediatric and adult examinations with various pathologies (n = 1248). Three deep learning model architectures (SegResNet, DynUNet, and SwinUNETR) from the Medical Open Network for Artificial Intelligence (MONAI) framework underwent training using native training (NT), relying solely on institutional datasets, and transfer learning (TL), incorporating pretraining on public datasets. For comparison, TotalSegmentator, a publicly available segmentation model, was applied to test data without further training. Segmentation performance was evaluated using mean Dice similarity coefficient (DSC), with manual segmentations as reference. RESULTS. For internal pediatric data, the DSC for TotalSegmentator, NT models, and TL models for normal liver was 0.953, 0.964-0.965, and 0.965-0.966, respectively; for normal spleen, 0.914, 0.942-0.945, and 0.937-0.945; for normal pancreas, 0.733, 0.774-0.785, and 0.775-0.786; and for pancreas with pancreatitis, 0.703, 0.590-0.640, and 0.667-0.711. For public pediatric data, the DSC for TotalSegmentator, NT models, and TL models for liver was 0.952, 0.871-0.908, and 0.941-0.946, respectively; for spleen, 0.905, 0.771-0.827, and 0.897-0.926; and for pancreas, 0.700, 0.577-0.648, and 0.693-0.736. For public primarily adult data, the DSC for TotalSegmentator, NT models, and TL models for liver was 0.991, 0.633-0.750, and 0.926-0.952, respectively; for spleen, 0.983, 0.569-0.604, and 0.923-0.947; and for pancreas, 0.909, 0.148-0.241, and 0.699-0.775. The DynUNet TL model was selected as the best-performing NT or TL model considering DSC values across organs and test datasets and was made available as an open-source MONAI bundle (https://github.com/cchmc-dll/pediatric_abdominal_segmentation_bundle.git). CONCLUSION. TL models trained on heterogeneous public datasets and fine-tuned using institutional pediatric data outperformed internal NT models and Total-Segmentator across internal and external pediatric test data. Segmentation performance was better in liver and spleen than in pancreas. CLINICAL IMPACT. The selected model may be used for various volumetry applications in pediatric imaging.

Automated Segmentation of MRI White Matter Hyperintensities in 8421 Patients with Acute Ischemic Stroke.

To date, only a few small studies have attempted deep learning-based automatic segmentation of white matter hyperintensity (WMH) lesions in patients with cerebral infarction; this issue is complicated because stroke-related lesions can obscure WMH borders. We developed and validated deep learning algorithms to segment WMH lesions accurately in patients with cerebral infarction using multisite data sets involving 8421 patients with acute ischemic stroke. We included 8421 patients with stroke from 9 centers in Korea. 2D UNet and squeeze-and-excitation (SE)-UNet models were trained using 2408 FLAIR MRIs from 3 hospitals and validated using 6013 FLAIR MRIs from 6 hospitals. WMH segmentation performance was assessed by calculating the Dice similarity coefficient (DSC), the correlation coefficient, and the concordance correlation coefficient compared with a human-segmented criterion standard. In addition, we obtained an uncertainty index that represents overall ambiguity in the voxel classification for WMH segmentation in each patient based on the Kullback-Leibler divergence. In the training data set, the mean age was 67.4 (SD, 13.0) years, and 60.4% were men. The mean (95% CI) DSCs for UNet in internal testing and external validation were, respectively, 0.659 (0.649-0.669) and 0.710 (0.707-0.714), which were slightly lower than the reliability between humans (DSC = 0.744; 95% CI, 0.738-0.751; P = .031). Compared with the UNet, the SE-UNet demonstrated better performance, achieving a mean DSC of 0.675 (95% CI, 0.666-0.685; P < .001) in the internal testing and 0.722 (95% CI, 0.719-0.726; P < .001) in the external validation; moreover, it achieved high DSC values (ranging from 0.672 to 0.744) across multiple validation data sets. We observed a significant correlation between WMH volumes that were segmented automatically and manually for the UNet (r = 0.917, P < .001), and it was even stronger for the SE-UNet (r = 0.933, P < .001). The SE-UNet also attained a high concordance correlation coefficient (ranging from 0.841 to 0.956) in the external test data sets. In addition, the uncertainty indices in most patients (86%) in the external data sets were <0.35, with an average DSC of 0.744 in these patients. We developed and validated deep learning algorithms to segment WMH in patients with acute cerebral infarction using the largest-ever MRI data sets. In addition, we showed that the uncertainty index can be used to identify cases in which automatic WMH segmentation is less accurate and requires human review.

Construction of Highly Accurate Machine Learning Potential Energy Surfaces for Excited-State Dynamics Simulations Based on Low-Level Data Sets.

Machine learning is capable of effectively predicting the potential energies of molecules in the presence of high-quality data sets. Its application in the construction of ground- and excited-state potential energy surfaces is attractive to accelerate nonadiabatic molecular dynamics simulations of photochemical reactions. Because of the huge computational cost of excited-state electronic structure calculations, the construction of a high-quality data set becomes a bottleneck. In the present work, we first built two data sets. One was obtained from surface hopping dynamics simulations at the semiempirical OM2/MRCI level. Another was extracted from the dynamics trajectories at the CASSCF level, which was reported previously. The ground- and excited-state potential energy surfaces of ethylene-bridged azobenzene at the CASSCF computational level were constructed based on the former low-level data set. Although non-neural network machine learning methods can achieve good or modest performance during the training process, only neural network models provide reliable predictions on the latter external test data set. The BPNN and SchNet combined with the Δ-ML scheme and the force term in the loss functions are recommended for dynamics simulations. Then, we performed excited-state dynamics simulations of the photoisomerization of ethylene-bridged azobenzene on machine learning potential energy surfaces. Compared with the lifetimes of the first excited state (S1) estimated at different computational levels, our results on the E isomer are in good agreement with the high-level estimation. However, the overestimation of the Z isomer is unimproved. It suggests that smaller errors during the training process do not necessarily translate to more accurate predictions on high-level potential energies or better performance on nonadiabatic dynamics simulations, at least in the present case.

Quantitative radiomics analysis of imaging features in adults and children Mycoplasma pneumonia.

This study aims to explore the value of clinical features, CT imaging signs, and radiomics features in differentiating between adults and children with Mycoplasma pneumonia and seeking quantitative radiomic representations of CT imaging signs. In a retrospective analysis of 981 cases of mycoplasmal pneumonia patients from November 2021 to December 2023, 590 internal data (adults:450, children: 140) randomly divided into a training set and a validation set with an 8:2 ratio and 391 external test data (adults:121; children:270) were included. Using univariate analysis, CT imaging signs and clinical features with significant differences (p < 0.05) were selected. After segmenting the lesion area on the CT image as the region of interest, 1,904 radiomic features were extracted. Then, Pearson correlation analysis (PCC) and the least absolute shrinkage and selection operator (LASSO) were used to select the radiomic features. Based on the selected features, multivariable logistic regression analysis was used to establish the clinical model, CT image model, radiomic model, and combined model. The predictive performance of each model was evaluated using ROC curves, AUC, sensitivity, specificity, accuracy, and precision. The AUC between each model was compared using the Delong test. Importantly, the radiomics features and quantitative and qualitative CT image features were analyzed using Pearson correlation analysis and analysis of variance, respectively. For the individual model, the radiomics model, which was built using 45 selected features, achieved the highest AUCs in the training set, validation set, and external test set, which were 0.995 (0.992, 0.998), 0.952 (0.921, 0.978), and 0.969 (0.953, 0.982), respectively. In all models, the combined model achieved the highest AUCs, which were 0.996 (0.993, 0.998), 0.972 (0.942, 0.995), and 0.986 (0.976, 0.993) in the training set, validation set, and test set, respectively. In addition, we selected 11 radiomics features and CT image features with a correlation coefficient r greater than 0.35. The combined model has good diagnostic performance for differentiating between adults and children with mycoplasmal pneumonia, and different CT imaging signs are quantitatively represented by radiomics.

Early prediction of gestational diabetes mellitus using maternal demographic and clinical risk factors

ObjectiveTo build and validate an early risk prediction model for gestational diabetes mellitus (GDM) based on first-trimester electronic medical records including maternal demographic and clinical risk factors.MethodsTo develop and validate a GDM prediction model, two datasets were used in this retrospective study. One included data of 14,015 pregnant women from Máxima Medical Center (MMC) in the Netherlands. The other was from an open-source database nuMoM2b including data of 10,038 nulliparous pregnant women, collected in the USA. Widely used maternal demographic and clinical risk factors were considered for modeling. A GDM prediction model based on elastic net logistic regression was trained from a subset of the MMC data. Internal validation was performed on the remaining MMC data to evaluate the model performance. For external validation, the prediction model was tested on an external test set from the nuMoM2b dataset.ResultsAn area under the receiver-operating-characteristic curve (AUC) of 0.81 was achieved for early prediction of GDM on the MMC test data, comparable to the performance reported in previous studies. While the performance markedly decreased to an AUC of 0.69 when testing the MMC-based model on the external nuMoM2b test data, close to the performance trained and tested on the nuMoM2b dataset only (AUC = 0.70).

Genome-wide 5-hydroxymethylcytosines in circulating cell-free DNA as noninvasive diagnostic markers for gastric cancer.

5-Hydroxymethylcytosine-enriched gene profiles and regions show tissue-specific and tumor specific. There is a potential value to explore cell-free DNA 5-hydroxymethylcytosine feature biomarkers for early gastric cancer detection. A matched case‒control study design with 50 gastric cancer patients and 50 controls was performed to sequence the different 5-hydroxymethylcytosine modification features of cell free DNA. Significantly differential 5-hydroxymethylcytosine modification genes were identified to construct a gastric cancer diagnostic model. Data set from GEO was used as an external testing set to test the robustness of the diagnostic model. Accounting for more than 90% of 5-hydroxymethylcytosine peaks were distributed in the gene body in both the gastric cancer and control groups. The diagnostic model was developed based on five different 5-hydroxymethylcytosine modification genes, FBXL7, PDE3A, TPO, SNTG2 and STXBP5. The model could effectively distinguish gastric cancer patients from controls in the training (AUC = 0.95, sensitivity = 88.6%, specificity = 94.3%), validation (AUC = 0.87, sensitivity = 73.3%, specificity = 93.3%) and testing (AUC = 0.90, sensitivity = 81.9%, specificity = 90.2%) sets. The risk scores of the controls from the model were significantly lower than those of gastric cancer patients in bothour own data (P < 0.001) and GEO external testing data (P < 0.001), and no significant difference between different TNM stage patients (P = 0.09 and 0.66). Furthermore, there was no significant difference between the healthy control and benign gastric disease patients in the testing set from GEO (P = 0.10). The characteristics of 5-hydroxymethylcytosine in cell free DNA are specific to gastric cancer patients, and the diagnostic model constructed by five genes' 5-hydroxymethylcytosine features could effectively identify gastric cancer patients.

Machine Learning Models for Prediction of Xenobiotic Chemicals with High Propensity to Transfer into Human Milk.

Breast milk serves as a vital source of essential nutrients for infants. However, human milk contamination via the transfer of environmental chemicals from maternal exposome is a significant concern for infant health. The milk to plasma concentration (M/P) ratio is a critical metric that quantifies the extent to which these chemicals transfer from maternal plasma into breast milk, impacting infant exposure. Machine learning-based predictive toxicology models can be valuable in predicting chemicals with a high propensity to transfer into human milk. To this end, we build such classification- and regression-based models by employing multiple machine learning algorithms and leveraging the largest curated data set, to date, of 375 chemicals with known milk-to-plasma concentration (M/P) ratios. Our support vector machine (SVM)-based classifier outperforms other models in terms of different performance metrics, when evaluated on both (internal) test data and an external test data set. Specifically, the SVM-based classifier on (internal) test data achieved a classification accuracy of 77.33%, a specificity of 84%, a sensitivity of 64%, and an F-score of 65.31%. When evaluated on an external test data set, our SVM-based classifier is found to be generalizable with a sensitivity of 77.78%. While we were able to build highly predictive classification models, our best regression models for predicting the M/P ratio of chemicals could achieve only moderate R2 values on the (internal) test data. As noted in the earlier literature, our study also highlights the challenges in developing accurate regression models for predicting the M/P ratio of xenobiotic chemicals. Overall, this study attests to the immense potential of predictive computational toxicology models in characterizing the myriad of chemicals in the human exposome.

TMTV-Net: fully automated total metabolic tumor volume segmentation in lymphoma PET/CT images - a multi-center generalizability analysis.

Total metabolic tumor volume (TMTV) segmentation has significant value enabling quantitative imaging biomarkers for lymphoma management. In this work, we tackle the challenging task of automated tumor delineation in lymphoma from PET/CT scans using a cascaded approach. Our study included 1418 2-[18F]FDG PET/CT scans from four different centers. The dataset was dividedinto 900 scans for development/validation/testing phasesand 518 for multi-center external testing. The former consisted of 450 lymphoma, lung cancer, and melanoma scans, along with 450 negative scans, while the latter consisted of lymphoma patients from different centers with diffuse large B cell, primary mediastinal large B cell, and classic Hodgkin lymphoma cases. Our approach involves resampling PET/CT images into different voxel sizes in the first step, followed by training multi-resolution 3D U-Nets on each resampled dataset using a fivefold cross-validation scheme. The models trained on different data splits were ensemble. After applying soft voting to the predicted masks, in the second step, we input the probability-averaged predictions, along with the input imaging data, into another 3D U-Net. Models were trained with semi-supervised loss. We additionally considered the effectiveness of using test time augmentation (TTA) to improve the segmentation performance after training. In addition to quantitative analysis including Dice score (DSC) and TMTV comparisons, the qualitative evaluation was also conducted by nuclear medicine physicians. Our cascaded soft-voting guided approach resulted in performance with an average DSC of 0.68 ± 0.12 for the internal test data from developmental dataset, and an average DSC of 0.66 ± 0.18 on the multi-site external data (n = 518), significantly outperforming (p < 0.001) state-of-the-art (SOTA) approaches including nnU-Net and SWIN UNETR. While TTA yielded enhanced performance gains for some of the comparator methods, its impact on our cascaded approach was found to be negligible (DSC: 0.66 ± 0.16). Our approach reliably quantified TMTV, with a correlation of 0.89 with the ground truth (p < 0.001). Furthermore, in terms of visual assessment, concordance between quantitative evaluations and clinician feedback was observed in the majority of cases. The average relative error (ARE) and the absolute error (AE) in TMTV prediction on external multi-centric dataset were ARE = 0.43 ± 0.54 and AE = 157.32 ± 378.12 (mL) for all the external test data (n = 518), and ARE = 0.30 ± 0.22 and AE = 82.05 ± 99.78 (mL) when the 10% outliers (n = 53) were excluded. TMTV-Net demonstrates strong performance and generalizability in TMTV segmentation across multi-site external datasets, encompassing various lymphoma subtypes. A negligible reduction of 2% in overall performance during testing on external data highlights robust model generalizability across different centers and cancer types, likely attributable to its training with resampled inputs. Our model is publicly available, allowing easy multi-site evaluation and generalizability analysis on datasets from different institutions.

PANCDR: precise medicine prediction using an adversarial network for cancer drug response.

Pharmacogenomics aims to provide personalized therapy to patients based on their genetic variability. However, accurate prediction of cancer drug response (CDR) is challenging due to genetic heterogeneity. Since clinical data are limited, most studies predicting drug response use preclinical data to train models. However, such models might not be generalizable to external clinical data due to differences between the preclinical and clinical datasets. In this study, a Precision Medicine Prediction using an Adversarial Network for Cancer Drug Response (PANCDR) model is proposed. PANCDR consists of two sub-models, an adversarial model and a CDR prediction model. The adversarial model reduces the gap between the preclinical and clinical datasets, while the CDR prediction model extracts features and predicts responses. PANCDR was trained using both preclinical data and unlabeled clinical data. Subsequently, it was tested on external clinical data, including The Cancer Genome Atlas and brain tumor patients. PANCDR outperformed other machine learning models in predicting external test data. Our results demonstrate the robustness of PANCDR and its potential in precision medicine by recommending patient-specific drug candidates. The PANCDR codes and data are available at https://github.com/DMCB-GIST/PANCDR.

Automated MRI liver segmentation for anatomical segmentation, liver volumetry, and the extraction of radiomics.

To develop and evaluate a deep convolutional neural network (DCNN) for automated liver segmentation, volumetry, and radiomic feature extraction on contrast-enhanced portal venous phase magnetic resonance imaging (MRI). This retrospective study included hepatocellular carcinoma patients from an institutional database with portal venous MRI. After manual segmentation, the data was randomly split into independent training, validation, and internal testing sets. From a collaborating institution, de-identified scans were used for external testing. The public LiverHccSeg dataset was used for further external validation. A 3D DCNN was trained to automatically segment the liver. Segmentation accuracy was quantified by theDice similarity coefficient (DSC) with respect to manual segmentation. A Mann-Whitney U test was used to compare the internal and external test sets. Agreement of volumetry and radiomic features was assessed using the intraclass correlation coefficient (ICC). In total, 470 patients met the inclusion criteria (63.9±8.2 years; 376 males) and 20 patients were used for external validation (41±12 years; 13 males). DSC segmentation accuracy of the DCNN was similarly high between the internal (0.97±0.01) and external (0.96±0.03) test sets (p=0.28) and demonstrated robust segmentation performance on public testing (0.93±0.03). Agreement of liver volumetry was satisfactory in the internal (ICC, 0.99), external (ICC, 0.97), and public (ICC, 0.85) test sets. Radiomic features demonstrated excellent agreement in the internal (mean ICC, 0.98±0.04), external (mean ICC, 0.94±0.10), and public (mean ICC, 0.91±0.09) datasets. Automated liver segmentation yields robust and generalizable segmentation performance on MRI data and can be used for volumetry and radiomic feature extraction. Liver volumetry, anatomic localization, and extraction of quantitative imaging biomarkers require accurate segmentation, but manual segmentation is time-consuming. A deep convolutional neural network demonstrates fast and accurate segmentation performance on T1-weighted portal venous MRI. • This deep convolutional neural network yields robust and generalizable liver segmentation performance on internal, external, and public testing data. • Automated liver volumetry demonstrated excellent agreement with manual volumetry. • Automated liver segmentations can be used for robust and reproducible radiomic feature extraction.

Prediction of H3K27M alteration in midline gliomas of the brain using radiomics: A multi-institute study.

Noninvasive prediction of H3K27M-altered Diffuse midline gliomas is important because of the involvement of deep locations and proximity to eloquent structures. We aim to predict H3K27M alteration in midline gliomas using radiomics features of T2W images. Radiomics features extracted from 124 subjects (69 H3K27M-altered/55 H3K27M-wild type). T2W images were resampled to 1 × 1 × 1mm3 voxel size, preprocessed, and normalized for artifact correction, intensity variations. The feature set was normalized and subjected to reduction by variance thresholding, correlation coefficient thresholding, and sequential feature selector. Adaptive synthesis oversampling technique was used to oversample the training data. Random forest classifier (RFC), Decision tree classifier (DTC), and K-nearest neighbors classifier (KNN) were trained over the training dataset and the performance was assessed over the internal test dataset and external test data set (52 subjects: 33 H3K27M-altered/19-H3K27M-wild type). DTC achieved a validation score of 77.33% (5-fold cross-validation) and an accuracy of 80.64%, 75% on internal and external test datasets. RFC achieved a validation score of 80.7% (5-fold cross-validation) an accuracy of 80.6%, and 73% on internal and external test datasets. DTC achieved a validation score of 78.67% (5-fold cross-validation) an accuracy of 80.64%, and 61.53% on internal and external test datasets. The accuracy score of DTC, RFC, and KNN on the internal test dataset was approximately 80% while on the external test dataset, DTC achieved 75% accuracy, RFC achieved 73% accuracy and KNN achieved 65.1% accuracy. H3K27M alteration is a potential immunotherapeutic marker and is associated with poor prognosis and radiomics features extracted from conventional T2W-images can help in identifying H3K27M-altered cases non-invasively with high precision.