Preoperative Contrast-Enhanced MRI in Differentiating Glioblastoma From Low-Grade Gliomas in The Cancer Imaging Archive Database: A Proof-of-Concept Study

PurposeThis study aimed to develop a repeatable MRI-based machine learning model to differentiate between low-grade gliomas (LGGs) and glioblastoma (GBM) and provide more clinical information to improve treatment decision-making.MethodsPreoperative MRIs of gliomas from The Cancer Imaging Archive (TCIA)–GBM/LGG database were selected. The tumor on contrast-enhanced MRI was segmented. Quantitative image features were extracted from the segmentations. A random forest classification algorithm was used to establish a model in the training set. In the test phase, a random forest model was tested using an external test set. Three radiologists reviewed the images for the external test set. The area under the receiver operating characteristic curve (AUC) was calculated. The AUCs of the radiomics model and radiologists were compared.ResultsThe random forest model was fitted using a training set consisting of 142 patients [mean age, 52 years ± 16 (standard deviation); 78 men] comprising 88 cases of GBM. The external test set included 25 patients (14 with GBM). Random forest analysis yielded an AUC of 1.00 [95% confidence interval (CI): 0.86–1.00]. The AUCs for the three readers were 0.92 (95% CI 0.74–0.99), 0.70 (95% CI 0.49–0.87), and 0.59 (95% CI 0.38–0.78). Statistical differences were only found between AUC and Reader 1 (1.00 vs. 0.92, respectively; p = 0.16).ConclusionAn MRI radiomics-based random forest model was proven useful in differentiating GBM from LGG and showed better diagnostic performance than that of two inexperienced radiologists.

Apparent Diffusion Coefficient-Based Convolutional Neural Network Model Can Be Better Than Sole Diffusion-Weighted Magnetic Resonance Imaging to Improve the Differentiation of Invasive Breast Cancer From Breast Ductal Carcinoma In Situ

Background and PurposeBreast ductal carcinoma in situ (DCIS) has no metastatic potential, and has better clinical outcomes compared with invasive breast cancer (IBC). Convolutional neural networks (CNNs) can adaptively extract features and may achieve higher efficiency in apparent diffusion coefficient (ADC)-based tumor invasion assessment. This study aimed to determine the feasibility of constructing an ADC-based CNN model to discriminate DCIS from IBC.MethodsThe study retrospectively enrolled 700 patients with primary breast cancer between March 2006 and June 2019 from our hospital, and randomly selected 560 patients as the training and validation sets (ratio of 3 to 1), and 140 patients as the internal test set. An independent external test set of 102 patients during July 2019 and May 2021 from a different scanner of our hospital was selected as the primary cohort using the same criteria. In each set, the status of tumor invasion was confirmed by pathologic examination. The CNN model was constructed to discriminate DCIS from IBC using the training and validation sets. The CNN model was evaluated using the internal and external tests, and compared with the discriminating performance using the mean ADC. The area under the curve (AUC), sensitivity, specificity, and accuracy were calculated to evaluate the performance of the previous model.ResultsThe AUCs of the ADC-based CNN model using the internal and external test sets were larger than those of the mean ADC (AUC: 0.977 vs. 0.866, P = 0.001; and 0.926 vs. 0.845, P = 0.096, respectively). Regarding the internal test set and external test set, the ADC-based CNN model yielded sensitivities of 0.893 and 0.873, specificities of 0.929 and 0.894, and accuracies of 0.907 and 0.902, respectively. Regarding the two test sets, the mean ADC showed sensitivities of 0.845 and 0.818, specificities of 0.821 and 0.829, and accuracies of 0.836 and 0.824, respectively. Using the ADC-based CNN model, the prediction only takes approximately one second for a single lesion.ConclusionThe ADC-based CNN model can improve the differentiation of IBC from DCIS with higher accuracy and less time.

An automated COVID-19 triage pipeline using artificial intelligence based on chest radiographs and clinical data

While COVID-19 diagnosis and prognosis artificial intelligence models exist, very few can be implemented for practical use given their high risk of bias. We aimed to develop a diagnosis model that addresses notable shortcomings of prior studies, integrating it into a fully automated triage pipeline that examines chest radiographs for the presence, severity, and progression of COVID-19 pneumonia. Scans were collected using the DICOM Image Analysis and Archive, a system that communicates with a hospital’s image repository. The authors collected over 6,500 non-public chest X-rays comprising diverse COVID-19 severities, along with radiology reports and RT-PCR data. The authors provisioned one internally held-out and two external test sets to assess model generalizability and compare performance to traditional radiologist interpretation. The pipeline was evaluated on a prospective cohort of 80 radiographs, reporting a 95% diagnostic accuracy. The study mitigates bias in AI model development and demonstrates the value of an end-to-end COVID-19 triage platform.

Deep Learning Classification of Active Tuberculosis Using Chest X-Rays: Efficacy of Transfer Learning and Generalization Performance of Cross-Population Datasets

Chest X-ray based diagnosis of active Tuberculosis (TB) is one of the oldest ubiquitous tests in medical practice. Artificial Intelligence (AI) based automated detection of abnormality in chest radiography is crucial in radiology workflow. Most deep convolutional neural networks (DCNN) for diagnosing TB by transfer learning from natural images and using the same dataset to evaluate the model performance and diagnostic accuracy. However, dataset shift is a known issue in predictive models in AI, which is unexplored. In this work, we fine-tuned, validated, and tested two benchmark architectures and utilized the transfer learning methodology to measure the diagnostic accuracy on cross-population datasets. We achieved remarkable calcification accuracy of 100% and area under the receiver operating characteristic (AUC) 1.000 [1.000 – 1.000] (with a sensitivity 0.985 [0.971 – 1.000] and a specificity of 0.986 [0.971 – 1.000]) on intramural test set, but significant drop in extramural test set. Accuracy on various extramural test sets varies 50% - 70%, AUC ranges 0.527 – 0.865 (sensitivity and specificity fluctuate 0.394 – 0.995 and 0.443 – 0.864 respectively). Diagnostic performance on the intramural test set observed in this study shows that DCNN can accurately classify active TB and normal chest radiographs, however the external test set shows DCNN is less likely to generalize well on models trained on specific population dataset.

Prediction of Protein pKa with Representation Learning

The behavior of proteins is closely related to the protonation states of the residues. Therefore, prediction and measurement of pKa are essential to understand the basic functions of proteins. In this work, we develop a new empirical scheme for protein pKa prediction that is based on deep representation learning. It combines machine learning with atomic environment vector (AEV) and learned quantum mechanical representation from ANI-2x neural network potential (J. Chem. Theory Comput. 2020, 16, 4192). The scheme requires only the coordinate information of a protein as the input and separately estimates the pKa for all five titratable amino acid types. The accuracy of the approach was analyzed with both cross-validation and an external test set of proteins. Obtained results were compared with the widely used empirical approach PROPKA. The new empirical model provides accuracy with MAEs below 0.5 for all amino acid types. It surpasses the accuracy of PROPKA and performs significantly better than the null model. Our model is also sensitive to the local conformational changes and molecular interactions.

Multi-Task Learning with Recurrent Neural Networks for ARDS Prediction using only EHR Data: Model Development and Validation Study (Preprint)

BACKGROUND Acute Respiratory Distress Syndrome (ARDS) is a condition that is often considered to have broad and subjective diagnostic criteria and is associated with significant mortality and morbidity. Early and accurate prediction of ARDS and related conditions such as hypoxemia and sepsis could allow timely administration of therapies, leading to improved patient outcomes. OBJECTIVE To perform an exploration of how multi-label classification in the clinical setting can take advantage of the underlying dependencies between ARDS and related conditions to improve early prediction of ARDS. METHODS The electronic health record dataset included 40,073 patient encounters from 7 hospitals from 4/20/2018 to 3/17/2021. A recurrent neural network (RNN) was trained using data from 5 hospitals, and external validation was conducted on data from 2 hospitals. In addition to ARDS, 12 target labels for related conditions such as sepsis, hypoxemia and Covid-19 were used to train the model to classify a total of 13 outputs. As a comparator, XGBoost models were developed for each of the 13 target labels. Model performance was assessed using the area under the receiver operating characteristic (AUROC). Heatmaps to visualize attention scores were generated to provide interpretability to the NNs. Finally, cluster analysis was performed to identify potential phenotypic subgroups of ARDS patients. RESULTS The single RNN model trained to classify 13 outputs outperformed the XGBoost model for ARDS prediction, achieving an AUROC of 0.842 on the external test sets. Models trained on an increasing number of tasks resulted in increasing performance. Earlier diagnosis of ARDS nearly doubled the rate of in-hospital survival. Cluster analysis revealed distinct ARDS subgroups, some of which had similar mortality rates but different clinical presentations. CONCLUSIONS The RNN model presented in this paper can be used as an early warning system to stratify patients who are at risk of developing one of the multiple risk outcomes, hence providing practitioners with means to take early action.

Automatic Pancreatic Cyst Lesion Segmentation on EUS Images Using a Deep-Learning Approach

The automatic segmentation of the pancreatic cyst lesion (PCL) is essential for the automated diagnosis of pancreatic cyst lesions on endoscopic ultrasonography (EUS) images. In this study, we proposed a deep-learning approach for PCL segmentation on EUS images. We employed the Attention U-Net model for automatic PCL segmentation. The Attention U-Net was compared with the Basic U-Net, Residual U-Net, and U-Net++ models. The Attention U-Net showed a better dice similarity coefficient (DSC) and intersection over union (IoU) scores than the other models on the internal test. Although the Basic U-Net showed a higher DSC and IoU scores on the external test than the Attention U-Net, there was no statistically significant difference. On the internal test of the cross-over study, the Attention U-Net showed the highest DSC and IoU scores. However, there was no significant difference between the Attention U-Net and Residual U-Net or between the Attention U-Net and U-Net++. On the external test of the cross-over study, all models showed no significant difference from each other. To the best of our knowledge, this is the first study implementing segmentation of PCL on EUS images using a deep-learning approach. Our experimental results show that a deep-learning approach can be applied successfully for PCL segmentation on EUS images.

Exploratory and machine learning analysis of the stability constants of HgII- triazene ligands complexes

Knowing stability constants for the complexes HgII with extracting ligands is very important from environmental and therapeutic standpoints. Since the selectivity of ligands can be stated by the stability constants of cation–ligand complexes, quantitative structure–property relationship (QSPR) investigations on binding constant of HgII complexes were done. Experimental data of the stability constants in ML2 complexation of HgII and synthesized triazene ligands were used to construct and develop QSPR models. Support vector machine (SVM) and multiple linear regression (MLR) have been employed to create the QSPR models. The final model showed squared correlation coefficient of 0.917 and the standard error of calibration (SEC) value of 0.141 log K units. The proposed model presented accurate prediction with the Leave-One-Out cross validation ( Q LOO 2 = 0.756) and validated using Y-randomization and external test set. Statistical results demonstrated that the proposed models had suitable goodness of fit, predictive ability, and robustness. The results revealed the importance of charge effects and topological properties of ligand in HgII - triazene complexation.

Classification of MR-Detected Additional Lesions in Patients With Breast Cancer Using a Combination of Radiomics Analysis and Machine Learning

ObjectiveThis study was conducted in order to investigate the feasibility of using radiomics analysis (RA) with machine learning algorithms based on breast magnetic resonance (MR) images for discriminating malignant from benign MR-detected additional lesions in patients with primary breast cancer.Materials and MethodsOne hundred seventy-four MR-detected additional lesions (benign, n = 86; malignancy, n = 88) from 158 patients with ipsilateral primary breast cancer from a tertiary medical center were included in this retrospective study. The entire data were randomly split to training (80%) and independent test sets (20%). In addition, 25 patients (benign, n = 21; malignancy, n = 15) from another tertiary medical center were included for the external test. Radiomics features that were extracted from three regions-of-interest (ROIs; intratumor, peritumor, combined) using fat-saturated T1-weighted images obtained by subtracting pre- from postcontrast images (SUB) and T2-weighted image (T2) were utilized to train the support vector machine for the binary classification. A decision tree method was utilized to build a classifier model using clinical imaging interpretation (CII) features assessed by radiologists. Area under the receiver operating characteristic curve (AUROC), accuracy, sensitivity, and specificity were used to compare the diagnostic performance.ResultsThe RA models trained using radiomics features from the intratumor-ROI showed comparable performance to the CII model (accuracy, AUROC: 73.3%, 69.6% for the SUB RA model; 70.0%, 75.1% for the T2 RA model; 73.3%, 72.0% for the CII model). The diagnostic performance increased when the radiomics and CII features were combined to build a fusion model. The fusion model that combines the CII features and radiomics features from multiparametric MRI data demonstrated the highest performance with an accuracy of 86.7% and an AUROC of 91.1%. The external test showed a similar pattern where the fusion models demonstrated higher levels of performance compared with the RA- or CII-only models. The accuracy and AUROC of the SUB+T2 RA+CII model in the external test were 80.6% and 91.4%, respectively.ConclusionOur study demonstrated the feasibility of using RA with machine learning approach based on multiparametric MRI for quantitatively characterizing MR-detected additional lesions. The fusion model demonstrated an improved diagnostic performance over the models trained with either RA or CII alone.

Intraoperative discrimination of native meningioma and dura mater by Raman spectroscopy

Meningiomas are among the most frequent tumors of the central nervous system. For a total resection, shown to decrease recurrences, it is paramount to reliably discriminate tumor tissue from normal dura mater intraoperatively. Raman spectroscopy (RS) is a non-destructive, label-free method for vibrational analysis of biochemical molecules. On the microscopic level, RS was already used to differentiate meningioma from dura mater. In this study we test its suitability for intraoperative macroscopic meningioma diagnostics. RS is applied to surgical specimen of intracranial meningiomas. The main purpose is the differentiation of tumor from normal dura mater, in order to potentially accelerate the diagnostic workflow. The collected meningioma and dura mater samples (n = 223 tissue samples from a total of 59 patients) are analyzed under untreated conditions using a new partially robotized RS acquisition system. Spectra (n = 1273) are combined with the according histopathological analysis for each sample. Based on this, a classifier is trained via machine learning. Our trained classifier separates meningioma and dura mater with a sensitivity of 96.06 $$\pm $$ ± 0.03% and a specificity of 95.44 $$\pm $$ ± 0.02% for internal fivefold cross validation and 100% and 93.97% if validated with an external test set. RS is an efficient method to discriminate meningioma from healthy dura mater in fresh tissue samples without additional processing or histopathological imaging. It is a quick and reliable complementary diagnostic tool to the neuropathological workflow and has potential for guided surgery. RS offers a safe way to examine unfixed surgical specimens in a perioperative setting.