Fully automated 3D machine learning model for HPV status characterization in oropharyngeal squamous cell carcinomas based on CT images.

BACKGROUND: Human papillomavirus (HPV) status plays a major role in predicting oropharyngeal squamous cell carcinoma (OPSCC) survival. This study assesses the accuracy of a fully automated 3D convolutional neural network (CNN) in predicting HPV status using CT images. METHODS: Pretreatment CT images from OPSCC patients were used to train a 3D DenseNet-121 model to predict HPV-p16 status. Performance was evaluated by the ROC Curve (AUC), sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and F1 score. RESULTS: The network achieved a mean AUC of 0.80 ± 0.06. The best-preforming fold had a sensitivity of 0.86 and specificity of 0.92 at the Youden's index. The PPV, NPV, and F1 scores are 0.97, 0.71, and 0.82, respectively. CONCLUSIONS: A fully automated CNN can characterize the HPV status of OPSCC patients with high sensitivity and specificity. Further refinement of this algorithm has the potential to provide a non-invasive tool to guide clinical management.

Multiparametric machine learning algorithm for human papillomavirus status and survival prediction in oropharyngeal cancer patients.

BACKGROUND: Human papillomavirus (HPV) status influences prognosis in oropharyngeal cancer (OPC). Identifying high-risk patients are critical to improving treatment. We aim to provide a noninvasive opportunity for managing OPC patients by training multiple machine learning pipelines to determine the best model for characterizing HPV status and survival. METHODS: Multi-parametric algorithms were designed using a 492 OPC patient database. HPV status incorporated age, sex, smoking/drinking habits, cancer subsite, TNM, and AJCC 7th edition staging. Survival considered HPV model inputs plus HPV status. Patients were split 4:1 training: testing. Algorithm efficacy was assessed through accuracy and area under the receiver operator characteristic curve (AUC). RESULTS: From 31 HPV status models, ensemble yielded 0.83 AUC and 78.7% accuracy. From 38 survival models, ensemble yielded 0.91 AUC and 87.7% accuracy. CONCLUSION: Results reinforce artificial intelligence's potential to use tumor imaging and patient characterizations for HPV status and outcome prediction. Utilizing these algorithms can optimize clinical guidance and patient care noninvasively.

A Fully Automated Deep Learning Network for Brain Tumor Segmentation.

We developed a fully automated method for brain tumor segmentation using deep learning; 285 brain tumor cases with multiparametric magnetic resonance images from the BraTS2018 data set were used. We designed 3 separate 3D-Dense-UNets to simplify the complex multiclass segmentation problem into individual binary-segmentation problems for each subcomponent. We implemented a 3-fold cross-validation to generalize the network's performance. The mean cross-validation Dice-scores for whole tumor (WT), tumor core (TC), and enhancing tumor (ET) segmentations were 0.92, 0.84, and 0.80, respectively. We then retrained the individual binary-segmentation networks using 265 of the 285 cases, with 20 cases held-out for testing. We also tested the network on 46 cases from the BraTS2017 validation data set, 66 cases from the BraTS2018 validation data set, and 52 cases from an independent clinical data set. The average Dice-scores for WT, TC, and ET were 0.90, 0.84, and 0.80, respectively, on the 20 held-out testing cases. The average Dice-scores for WT, TC, and ET on the BraTS2017 validation data set, the BraTS2018 validation data set, and the clinical data set were as follows: 0.90, 0.80, and 0.78; 0.90, 0.82, and 0.80; and 0.85, 0.80, and 0.77, respectively. A fully automated deep learning method was developed to segment brain tumors into their subcomponents, which achieved high prediction accuracy on the BraTS data set and on the independent clinical data set. This method is promising for implementation into a clinical workflow.

A novel fully automated MRI-based deep-learning method for classification of IDH mutation status in brain gliomas.

BACKGROUND: Isocitrate dehydrogenase (IDH) mutation status has emerged as an important prognostic marker in gliomas. Currently, reliable IDH mutation determination requires invasive surgical procedures. The purpose of this study was to develop a highly accurate, MRI-based, voxelwise deep-learning IDH classification network using T2-weighted (T2w) MR images and compare its performance to a multicontrast network. METHODS: Multiparametric brain MRI data and corresponding genomic information were obtained for 214 subjects (94 IDH-mutated, 120 IDH wild-type) from The Cancer Imaging Archive and The Cancer Genome Atlas. Two separate networks were developed, including a T2w image-only network (T2-net) and a multicontrast (T2w, fluid attenuated inversion recovery, and T1 postcontrast) network (TS-net) to perform IDH classification and simultaneous single label tumor segmentation. The networks were trained using 3D Dense-UNets. Three-fold cross-validation was performed to generalize the networks' performance. Receiver operating characteristic analysis was also performed. Dice scores were computed to determine tumor segmentation accuracy. RESULTS: T2-net demonstrated a mean cross-validation accuracy of 97.14% ± 0.04 in predicting IDH mutation status, with a sensitivity of 0.97 ± 0.03, specificity of 0.98 ± 0.01, and an area under the curve (AUC) of 0.98 ± 0.01. TS-net achieved a mean cross-validation accuracy of 97.12% ± 0.09, with a sensitivity of 0.98 ± 0.02, specificity of 0.97 ± 0.001, and an AUC of 0.99 ± 0.01. The mean whole tumor segmentation Dice scores were 0.85 ± 0.009 for T2-net and 0.89 ± 0.006 for TS-net. CONCLUSION: We demonstrate high IDH classification accuracy using only T2-weighted MR images. This represents an important milestone toward clinical translation.

Classification of brain tumor isocitrate dehydrogenase status using MRI and deep learning.

Isocitrate dehydrogenase (IDH) mutation status is an important marker in glioma diagnosis and therapy. We propose an automated pipeline for noninvasively predicting IDH status using deep learning and T2-weighted (T2w) magnetic resonance (MR) images with minimal preprocessing (N4 bias correction and normalization to zero mean and unit variance). T2w MR images and genomic data were obtained from The Cancer Imaging Archive dataset for 260 subjects (120 high-grade and 140 low-grade gliomas). A fully automated two-dimensional densely connected model was trained to classify IDH mutation status on 208 subjects and tested on another held-out set of 52 subjects using fivefold cross validation. Data leakage was avoided by ensuring subject separation during the slice-wise randomization. Mean classification accuracy of 90.5% was achieved for each axial slice in predicting the three classes of no tumor, IDH mutated, and IDH wild type. Test accuracy of 83.8% was achieved in predicting IDH mutation status for individual subjects on the test dataset of 52 subjects. We demonstrate a deep learning method to predict IDH mutation status using T2w MRI alone. Radiologic imaging studies using deep learning methods must address data leakage (subject duplication) in the randomization process to avoid upward bias in the reported classification accuracy.

Measuring Intraventricular Pressure Using Ultrasound Elastography.

OBJECTIVES: Intraventricular pressure (IVP) is one of the most important measurements for evaluating cardiac function, but this measurement is not currently easily assessable in the clinic. The primary reason for this is the absence of a noninvasive technique for measuring IVP. In this study, we investigate the relationship between IVP and dynamic myocardial stiffness measured by shear wave elasticity imaging (SWEI) and assess the feasibility of measuring IVP using SWEI. METHODS: In 8 isolated working rabbit hearts, IVP was recorded in the left ventricle using a pressure catheter. Simultaneously, myocardial stiffness was recorded by SWEI. Using the peak values for IVP and SWEI measured stiffness, SWEI measurements were calibrated and converted to IVP. RESULTS: A linear relationship with zero intercept was observed between IVP and SWEI, with the average slope of 0.318 kPa/mm Hg, R2 = 0.89. Using one point on the IVP/SWEI curve, SWEI measurements were converted to IVP. Estimated pressure using SWEI and IVP were linearly correlated with the slope of 0.95, R2 = 0.88 (mean end diastolic pressure by pressure catheter = 12.716 mm Hg and by SWEI=14.726 mm Hg), indicating the near equivalence of the 2 measurements. CONCLUSION: We have shown that SWEI measurements are linearly related to IVP; therefore, pressure-based indices could potentially be derived from SWEI ultrasound elastography. The feasibility of using SWEI to estimate IVP with a single point calibration was also shown in this study.

The mechanical and hemodynamic effects of left ventricular pacing in heart failure with preserved ejection fraction and left bundle branch block.

Cardiac resynchronization therapy (CRT) has revolutionized the care of patients with heart failure with reduced ejection fraction (HFrEF) and left bundle branch block (LBBB); some hypothesize that electrical resynchronization may also benefit patients with heart failure with preserved ejection fraction (HFpEF) and LBBB. We assessed the acute hemodynamic and mechanical impact of temporary LV pacing in 2 patients with HFpEF and LBBB and a "classic" pattern of echocardiographic dyssynchrony. LV pacing facilitated electrical resynchronization with acute resolution of mechanical dyssynchrony and improvements in invasively and non-invasively measured global cardiac function, due in part to shortening of the isovolumetric contraction period.

Assessment of Diastolic Function Using Ultrasound Elastography.

Shear wave elasticity imaging (SWEI) is a novel ultrasound elastography technique for assessing tissue stiffness. In this study, we investigate the potential of SWEI for providing diastolic functional assessment. In 11 isolated rabbit hearts, pressure-volume (PV) measurements were recorded simultaneously with SWEI recordings from the left ventricle free wall before and after induction of global ischemia. PV-based end diastolic stiffness increased by 100% after ischemia (p <0.05), and SWEI stiffness showed an increase of 103% (p <0.05). The relaxation time constant (τ) before and after ischemia derived from pressure and SWEI curves showed increases of 79% and 76%, respectively (p <0.05). A linear regression between pressure-derived and SWEI-based (τ) showed a slope of 1.164 with R2 = 0.80, indicating the near equivalence of the two assessments. SWEI can be used to derive (τ) values and myocardial end diastolic stiffness. In global conditions, these measurements are consistent with PV measurements of diastolic function.

Quantifying Myocardial Contractility Changes Using Ultrasound-Based Shear Wave Elastography.

BACKGROUND: Myocardial contractility, a significant determinant of cardiac function, is valuable for diagnosis and evaluation of treatment in cardiovascular disorders including heart failure. Shear wave elasticity imaging (SWEI) is a newly developed ultrasound-based elastographic technique that can directly assess the stiffness of cardiac tissue. The aim of this study was to verify the ability of this technique to quantify contractility changes in the myocardium. METHODS: In 12 isolated rabbit hearts, SWEI measurements were made of systolic stiffness at five different coronary perfusion pressures from 0 to 92 mm Hg. The changes in coronary perfusion were used to induce acute stepwise reversible changes in cardiac contractility via the Gregg effect. The Gregg effect is the dependency of contractility on coronary perfusion. In four of the hearts, the measurements were repeated after delivery of gadolinium, which is known to block the Gregg effect. RESULTS: Systolic stiffness measured by SWEI changed linearly with coronary perfusion pressure, with a slope of 0.27 kPa/mm Hg (mean of 95% CI, R2 = 0.73). As expected, the change in contractility due to the Gregg effect was blocked by gadolinium, with a significant reduction of the slope to 0.08 kPa/mm Hg. CONCLUSIONS: SWEI measurements of systolic stiffness provide an index of contractility in the unloaded isolated rabbit heart. Although this study was done under ideal imaging conditions and with nonphysiologic loading conditions, it reinforces the concept that this ultrasound technique has the potential to provide a direct and noninvasive index of cardiac contractility.

A Comparison of Acoustic Radiation Force-Derived Indices of Cardiac Function in the Langendorff Perfused Rabbit Heart.

In the past decade, there has been an increased interest in characterizing cardiac tissue mechanics utilizing newly developed ultrasound-based elastography techniques. These methods excite the tissue mechanically and track the response. Two frequently used methods, acoustic radiation force impulse (ARFI) and shear-wave elasticity imaging (SWEI), have been considered qualitative and quantitative techniques providing relative and absolute measures of tissue stiffness, respectively. Depending on imaging conditions, it is desirable to identify indices of cardiac function that could be measured by ARFI and SWEI and to characterize the relationship between the measures. In this study, we have compared two indices (i.e., relaxation time constant used for diastolic dysfunction assessment and systolic/diastolic stiffness ratio) measured nearly simultaneously by M-mode ARFI and SWEI techniques. We additionally correlated ARFI-measured inverse displacements with SWEI-measured values of the shear modulus of stiffness. For the eight animals studied, the average relaxation time constant ( τ) measured by ARFI and SWEI were ([Formula: see text]) and ([Formula: see text]), respectively ([Formula: see text]). Average systolic/diastolic stiffness ratios for ARFI and SWEI measurements were 6.01±1.37 and 7.12±3.24, respectively ([Formula: see text]). Shear modulus of stiffness (SWEI) was linearly related to inverse displacement values (ARFI) with a 95% CI for the slope of 0.010-0.011 [Formula: see text] ( R(2)=0.73). In conclusion, the relaxation time constant and the systolic/diastolic stiffness ratio were calculated with good agreement between the ARFI- and SWEI-derived measurements. ARFI relative and SWEI absolute stiffness measurements were linearly related with varying slopes based on imaging conditions and subject tissue properties.

The Evolution of Tissue Stiffness at Radiofrequency Ablation Sites During Lesion Formation and in the Peri-Ablation Period.

INTRODUCTION: Elastography imaging can provide radiofrequency ablation (RFA) lesion assessment due to tissue stiffening at the ablation site. An important aspect of assessment is the spatial and temporal stability of the region of stiffness increase in the peri-ablation period. The aim of this study was to use 2 ultrasound-based elastography techniques, shear wave elasticity imaging (SWEI) and acoustic radiation force impulse (ARFI) imaging, to monitor the evolution of tissue stiffness at ablation sites in the 30 minutes following lesion creation. METHODS AND RESULTS: In 6 canine subjects, SWEI measurements and 2-D ARFI images were acquired at 6 ventricular endocardial RFA sites before, during, and for 30 minutes postablation. An immediate increase in tissue stiffness was detected during RFA, and the area of the postablation region of stiffness increase (RoSI) as well as the relative stiffness at the RoSI center was stable approximately 2 minutes after ablation. Of note is the observation that relative stiffness in the region adjacent to the RoSI increased slightly during the first 15 minutes, consistent with local fluid displacement or edema. The magnitude of this increase, ∼0.5-fold from baseline, was significantly less than the magnitude of the stiffness increase directly inside the RoSI, which was greater than 3-fold from baseline. CONCLUSIONS: Ultrasound-based SWEI and ARFI imaging detected an immediate increase in tissue stiffness during RFA, and the stability and magnitude of the stiffness change suggest that consistent elasticity-based lesion assessment is possible 2 minutes after and for at least 30 minutes following ablation.

Ultrasound shear wave elasticity imaging quantifies coronary perfusion pressure effect on cardiac compliance.

Diastolic heart failure (DHF) is a major source of cardiac related morbidity and mortality in the world today. A major contributor to, or indicator of DHF is a change in cardiac compliance. Currently, there is no accepted clinical method to evaluate the compliance of cardiac tissue in diastolic dysfunction. Shear wave elasticity imaging (SWEI) is a novel ultrasound-based elastography technique that provides a measure of tissue stiffness. Coronary perfusion pressure affects cardiac stiffness during diastole; we sought to characterize the relationship between these two parameters using the SWEI technique. In this work, we demonstrate how changes in coronary perfusion pressure are reflected in a local SWEI measurement of stiffness during diastole. Eight Langendorff perfused isolated rabbit hearts were used in this study. Coronary perfusion pressure was changed in a randomized order (0-90 mmHg range) and SWEI measurements were recorded during diastole with each change. Coronary perfusion pressure and the SWEI measurement of stiffness had a positive linear correlation with the 95% confidence interval (CI) for the slope of 0.009-0.011 m/s/mmHg ( R(2) = 0.88 ). Furthermore, shear modulus was linearly correlated to the coronary perfusion pressure with the 95% CI of this slope of 0.035-0.042 kPa/mmHg ( R(2) = 0.83). In conclusion, diastolic SWEI measurements of stiffness can be used to characterize factors affecting cardiac compliance specifically the mechanical interaction (cross-talk) between perfusion pressure in the coronary vasculature and cardiac muscle. This relationship was found to be linear over the range of pressures tested.

The effect of acute coronary perfusion change on cardiac function measured by Shear Wave Elasticity Imaging.

The possibility of measuring cardiac function noninvasively has generated wide interest in elastography imaging techniques. Shear Wave Elasticity Imaging (SWEI) is an ultrasound-based elastography technique used to measure stiffness of tissues. While this technique has been studied extensively in static homogenous tissues such as liver, breast or prostate, there is still a significant need to study its capabilities to measure cardiac stiffness and function. In this research, we have studied the potential of SWEI to evaluate the coronary perfusion pressure effect on systolic and diastolic stiffness referred to as elastance and compliance of the heart. Five isolated rabbit hearts were used in this study in a Langendorff preparation. SWEI measurements of stiffness were recorded in two steps. In the first step, coronary perfusion was set to normal and then was reduced to half-normal. After 40 minutes of half-normal perfusion, it was returned to normal perfusion for the second step. SWEI velocity decreased from 6.003 m/s to 4.713 m/s in systole and from 1.948 m/s to 1.507 m/s in diastole in the first step. During the second step raising the perfusion to normal, SWEI stiffness showed an increase from 3.760 m/s to 5.468 m/s in systole and from 1.678 m/s to 2.156 m/s during diastole. Our results show that SWEI measurements of stiffness can characterize the cross talk between coronary perfusion and cardiac stiffness and also has the potential to measure compliance and elastance of the heart in systole and diastole.