Predicting EGFR Status After Radical Nephrectomy or Partial Nephrectomy for Renal Cell Carcinoma on CT Using a Self-attention-based Model: Variable Vision Transformer (vViT).

OBJECTIVE: To assess the effectiveness of the vViT model for predicting postoperative renal function decline by leveraging clinical data, medical images, and image-derived features; and to identify the most dominant factor influencing this prediction. MATERIALS AND METHODS: We developed two models, eGFR10 and eGFR20, to identify patients with a postoperative reduction in eGFR of more than 10 and more than 20, respectively, among renal cell carcinoma patients. The eGFR10 model was trained on 75 patients and tested on 27, while the eGFR20 model was trained on 77 patients and tested on 24. The vViT model inputs included class token, patient characteristics (age, sex, BMI), comorbidities (peripheral vascular disease, diabetes, liver disease), habits (smoking, alcohol), surgical details (ischemia time, blood loss, type and procedure of surgery, approach, operative time), radiomics, and tumor and kidney imaging. We used permutation feature importance to evaluate each sector's contribution. The performance of vViT was compared with CNN models, including VGG16, ResNet50, and DenseNet121, using McNemar and DeLong tests. RESULTS: The eGFR10 model achieved an accuracy of 0.741 and an AUC-ROC of 0.692, while the eGFR20 model attained an accuracy of 0.792 and an AUC-ROC of 0.812. The surgical and radiomics sectors were the most influential in both models. The vViT had higher accuracy and AUC-ROC than VGG16 and ResNet50, and higher AUC-ROC than DenseNet121 (p < 0.05). Specifically, the vViT did not have a statistically different AUC-ROC compared to VGG16 (p = 1.0) and ResNet50 (p = 0.7) but had a statistically different AUC-ROC compared to DenseNet121 (p = 0.87) for the eGFR10 model. For the eGFR20 model, the vViT did not have a statistically different AUC-ROC compared to VGG16 (p = 0.72), ResNet50 (p = 0.88), and DenseNet121 (p = 0.64). CONCLUSION: The vViT model, a transformer-based approach for multimodal data, shows promise for preoperative CT-based prediction of eGFR status in patients with renal cell carcinoma.

Inconsistency between Human Observation and Deep Learning Models: Assessing Validity of Postmortem Computed Tomography Diagnosis of Drowning.

Drowning diagnosis is a complicated process in the autopsy, even with the assistance of autopsy imaging and the on-site information from where the body was found. Previous studies have developed well-performed deep learning (DL) models for drowning diagnosis. However, the validity of the DL models was not assessed, raising doubts about whether the learned features accurately represented the medical findings observed by human experts. In this paper, we assessed the medical validity of DL models that had achieved high classification performance for drowning diagnosis. This retrospective study included autopsy cases aged 8-91 years who underwent postmortem computed tomography between 2012 and 2021 (153 drowning and 160 non-drowning cases). We first trained three deep learning models from a previous work and generated saliency maps that highlight important features in the input. To assess the validity of models, pixel-level annotations were created by four radiological technologists and further quantitatively compared with the saliency maps. All the three models demonstrated high classification performance with areas under the receiver operating characteristic curves of 0.94, 0.97, and 0.98, respectively. On the other hand, the assessment results revealed unexpected inconsistency between annotations and models' saliency maps. In fact, each model had, respectively, around 30%, 40%, and 80% of irrelevant areas in the saliency maps, suggesting the predictions of the DL models might be unreliable. The result alerts us in the careful assessment of DL tools, even those with high classification performance.

How intra-source imbalanced datasets impact the performance of deep learning for COVID-19 diagnosis using chest X-ray images.

Over the past decade, the use of deep learning has been widely increasing in the medical image diagnosis field. Deep learning-based methods' (DLMs) performance strongly relies on training data. Therefore, researchers often focus on collecting as much data as possible from different medical facilities or developing approaches to avoid the impact of inter-category imbalance (ICI), which means a difference in data quantity among categories. However, due to the ICI within each medical facility, medical data are often isolated and acquired in different settings among medical facilities, known as the issue of intra-source imbalance (ISI) characteristic. This imbalance also impacts the performance of DLMs but receives negligible attention. In this study, we study the impact of the ISI on DLMs by comparison of the version of a deep learning model that was trained separately by an intra-source imbalanced chest X-ray (CXR) dataset and an intra-source balanced CXR dataset for COVID-19 diagnosis. The finding is that using the intra-source imbalanced dataset causes a serious training bias, although the dataset has a good inter-category balance. In contrast, the deep learning model performed a reliable diagnosis when trained on the intra-source balanced dataset. Therefore, our study reports clear evidence that the intra-source balance is vital for training data to minimize the risk of poor performance of DLMs.

Risk Analysis of Breast Cancer by Using Bilateral Mammographic Density Differences: A Case-Control Study.

The identification of risk factors helps radiologists assess the risk of breast cancer. Quantitative factors such as age and mammographic density are established risk factors for breast cancer. Asymmetric breast findings are frequently encountered during diagnostic mammography. The asymmetric area may indicate a developing mass in the early stage, causing a difference in mammographic density between the left and right sides. Therefore, this paper aims to propose a quantitative parameter named bilateral mammographic density difference (BMDD) for the quantification of breast asymmetry and to verify BMDD as a risk factor for breast cancer. To quantitatively evaluate breast asymmetry, we developed a semi-automatic method to estimate mammographic densities and calculate BMDD as the absolute difference between the left and right mammographic densities. And then, a retrospective case-control study, covering the period from July 2006 to October 2014, was conducted to analyse breast cancer risk in association with BMDD. The study included 364 women diagnosed with breast cancer and 364 matched control patients. As a result, a significant difference in BMDD was found between cases and controls (P < 0.001) and the case-control study demonstrated that women with BMDD > 10% had a 2.4-fold higher risk of breast cancer (odds ratio, 2.4; 95% confidence interval, 1.3-4.5) than women with BMDD ≤ 10%. In addition, we also demonstrated the positive association between BMDD and breast cancer risk among the subgroups with different ages and the Breast Imaging Reporting and Data System (BI-RADS) mammographic density categories. This study demonstrated that BMDD could be a potential risk factor for breast cancer.

Comprehensive evaluation of machine learning algorithms for predicting sleep-wake conditions and differentiating between the wake conditions before and after sleep during pregnancy based on heart rate variability.

Introduction: Perinatal women tend to have difficulties with sleep along with autonomic characteristics. This study aimed to identify a machine learning algorithm capable of achieving high accuracy in predicting sleep-wake conditions and differentiating between the wake conditions before and after sleep during pregnancy based on heart rate variability (HRV). Methods: Nine HRV indicators (features) and sleep-wake conditions of 154 pregnant women were measured for 1 week, from the 23rd to the 32nd weeks of pregnancy. Ten machine learning and three deep learning methods were applied to predict three types of sleep-wake conditions (wake, shallow sleep, and deep sleep). In addition, the prediction of four conditions, in which the wake conditions before and after sleep were differentiated-shallow sleep, deep sleep, and the two types of wake conditions-was also tested. Results and Discussion: In the test for predicting three types of sleep-wake conditions, most of the algorithms, except for Naïve Bayes, showed higher areas under the curve (AUCs; 0.82-0.88) and accuracy (0.78-0.81). The test using four types of sleep-wake conditions with differentiation between the wake conditions before and after sleep also resulted in successful prediction by the gated recurrent unit with the highest AUC (0.86) and accuracy (0.79). Among the nine features, seven made major contributions to predicting sleep-wake conditions. Among the seven features, "the number of interval differences of successive RR intervals greater than 50 ms (NN50)" and "the proportion dividing NN50 by the total number of RR intervals (pNN50)" were useful to predict sleep-wake conditions unique to pregnancy. These findings suggest alterations in the vagal tone system specific to pregnancy.

Deep Learning-Based Diagnosis of Fatal Hypothermia Using Post-Mortem Computed Tomography.

In forensic medicine, fatal hypothermia diagnosis is not always easy because findings are not specific, especially if traumatized. Post-mortem computed tomography (PMCT) is a useful adjunct to the cause-of-death diagnosis and some qualitative image character analysis, such as diffuse hyperaeration with decreased vascularity or pulmonary emphysema, have also been utilized for fatal hypothermia. However, it is challenging for inexperienced forensic pathologists to recognize the subtle differences of fatal hypothermia in PMCT images. In this study, we developed a deep learning-based diagnosis system for fatal hypothermia and explored the possibility of being an alternative diagnostic for forensic pathologists. An in-house dataset of forensic autopsy proven samples was used for the development and performance evaluation of the deep learning system. We used the area under the receiver operating characteristic curve (AUC) of the system for evaluation, and a human-expert comparable AUC value of 0.905, sensitivity of 0.948, and specificity of 0.741 were achieved. The experimental results clearly demonstrated the usefulness and feasibility of the deep learning system for fatal hypothermia diagnosis.

Comments on "Convergence Analysis of Adaptive Exponential Functional Link Network".

This article is to comment on the derivation of the weight-update stability of in-parameter-linear nonlinear learning system with the gradient descent learning rule in the above article. Our comments are not to disqualify the commented article's whole contribution; however, the issues should be pointed out to avoid their proliferation.

Letter on Convergence of In-Parameter-Linear Nonlinear Neural Architectures With Gradient Learnings.

This letter summarizes and proves the concept of bounded-input bounded-state (BIBS) stability for weight convergence of a broad family of in-parameter-linear nonlinear neural architectures (IPLNAs) as it generally applies to a broad family of incremental gradient learning algorithms. A practical BIBS convergence condition results from the derived proofs for every individual learning point or batches for real-time applications.

A 2.5D Deep Learning-Based Method for Drowning Diagnosis Using Post-Mortem Computed Tomography.

It is challenging to diagnose drowning in autopsy even with the help of post-mortem multi-slice computed tomography (MSCT) due to the complex pathophysiology and the shortage of forensic specialists equipped with radiology knowledge. Therefore, a computer-aided diagnosis (CAD) system was developed to help with diagnosis. Most deep learning-based CAD systems only utilize 2D information, which is proper for 2D data such as chest X-ray images. However, 3D information should also be considered for 3D data like CT. Conventional 3D methods require a huge amount of data and computational cost when using 3D methods. In this article, we proposed a 2.5D method that converts 3D data into 2D images to train 2D deep learning models for drowning diagnosis. The key point of this 2.5D method is that it uses a subset to represent the whole case, covering this case as much as possible while avoiding other repetitive information. To evaluate the effectiveness of the proposed method, conventional 2D, previous 2.5D, and 3D deep learning-based methods were tested using an MSCT dataset obtained from Tohoku university. Then, to provide explainable diagnosis results, a visualization method called Gradient-weighted Class Activation Mapping was employed to visualize features relevant to drowning in CT images. Results on drowning diagnosis showed that our proposed method achieved the best performance compared to other 2D, 2.5D, and 3D methods. The visual assessment also demonstrated that our method could find the saliency regions corresponding to drowning.

Diagnosing Drowning in Postmortem CT Images Using Artificial Intelligence.

Imaging features of the lung in postmortem computed tomography (CT) scans have been reported in drowning cases. However, it is difficult for forensic pathologists with limited experience to distinguish subtle differences in CT images. In this study, artificial intelligence (AI) with deep learning capability was used to diagnose drowning in postmortem CT images, and its performance was evaluated. The samples consisted of high-resolution CT images of the chest of 153 drowned and 160 non-drowned bodies captured by an 8- or 64-row multislice CT system. The images were captured with an image slice thickness of 1.0 mm and spacing of 30 mm, and 28 images were typically captured. A modified AlexNet was used as the AI architecture. The output result was the drowning probability for each component image. To evaluate the performance of the proposed model, the area under the receiver operating characteristic curve (AUC) was analyzed, and the AUC value of 0.95 was obtained. This indicates that the proposed AI architecture is a useful and powerful complementary testing approach for diagnosing drowning in postmortem CT images. Notably, the accuracy was 81% (62/77) for cases in which resuscitation was performed, and 92% (216/236) for cases in which resuscitation was not attempted. Therefore, the proposed AI method should not be used to diagnose the cause of death when aggressive cardiopulmonary resuscitation was performed. Additionally, because honeycomb lungs are likely to exhibit different morphologies, emphysema cases should also be treated with caution when the proposed AI method is used to diagnose drowning.

[A Review of Current Knowledge for X-ray Energy in CT: Practical Guide for CT Technologist].

In computed tomography (CT) systems, the optimal X-ray energy in imaging depends on the material composition and the subject size. Among the parameters related to the X-ray energy, we can arbitrarily change only the tube voltage. For years, the tube voltage has often been set at 120 kVp. However, since about 2000, there has been an increasing interest in reducing radiation dose, and it has led to the publication of various reports on low tube voltage. Furthermore, with the spread of dual-energy CT, virtual monochromatic X-ray images are widely used since the contrast can be adjusted by selecting the optional energy. Therefore, because of the renewed interest in X-ray energy in CT imaging, the issue of energy and imaging needs to be summarized. In this article, we describe the basics of physical characteristics of X-ray attenuation with materials and its influence on the process of CT imaging. Moreover, the relationship between X-ray energy and CT imaging is discussed for clinical applications.

Manganese Dynamics in Mouse Brain After Systemic MnCl2 Administration for Activation-Induced Manganese-Enhanced MRI.

Activation-induced manganese-enhanced MRI (AIM-MRI) is an attractive tool for non-invasively mapping whole brain activities. Manganese ions (Mn2+) enter and accumulate in active neurons via calcium channels. Mn2+ shortens the longitudinal relaxation time (T1) of H+, and the longitudinal relaxation rate R1 (1/T1) is proportional to Mn2+ concentration. Thus, AIM-MRI can map neural activities throughout the brain by assessing the R1 map. However, AIM-MRI is still not widely used, partially due to insufficient information regarding Mn2+ dynamics in the brain. To resolve this issue, we conducted a longitudinal study looking at manganese dynamics after systemic administration of MnCl2 by AIM-MRI with quantitative analysis. In the ventricle, Mn2+ increased rapidly within 1 h, remained high for 3 h, and returned to near control levels by 24 h after administration. Microdialysis showed that extracellular Mn returned to control levels by 4 h after administration, indicating a high concentration of extracellular Mn2+ lasts at least about 3 h after administration. In the brain parenchyma, Mn2+ increased slowly, peaked 24-48 h after administration, and returned to control level by 5 days after a single administration and by 2 weeks after a double administration with a 24-h interval. These time courses suggest that AIM-MRI records neural activity 1-3 h after MnCl2 administration, an appropriate timing of the MRI scan is in the range of 24-48 h following systemic administration, and at least an interval of 5 days or a couple of weeks for single or double administrations, respectively, is needed for a repeat AIM-MRI experiment.

Heart Rate Information-Based Machine Learning Prediction of Emotions Among Pregnant Women.

In this study, the extent to which different emotions of pregnant women can be predicted based on heart rate-relevant information as indicators of autonomic nervous system functioning was explored using various machine learning algorithms. Nine heart rate-relevant autonomic system indicators, including the coefficient of variation R-R interval (CVRR), standard deviation of all NN intervals (SDNN), and square root of the mean squared differences of successive NN intervals (RMSSD), were measured using a heart rate monitor (MyBeat) and four different emotions including "happy," as a positive emotion and "anxiety," "sad," "frustrated," as negative emotions were self-recorded on a smartphone application, during 1 week starting from 23rd to 32nd weeks of pregnancy from 85 pregnant women. The k-nearest neighbor (k-NN), support vector machine (SVM), logistic regression (LR), random forest (RF), naïve bayes (NB), decision tree (DT), gradient boosting trees (GBT), stochastic gradient descent (SGD), extreme gradient boosting (XGBoost), and artificial neural network (ANN) machine learning methods were applied to predict the four different emotions based on the heart rate-relevant information. To predict four different emotions, RF also showed a modest area under the receiver operating characteristic curve (AUC-ROC) of 0.70. CVRR, RMSSD, SDNN, high frequency (HF), and low frequency (LF) mostly contributed to the predictions. GBT displayed the second highest AUC (0.69). Comprehensive analyses revealed the benefits of the prediction accuracy of the RF and GBT methods and were beneficial to establish models to predict emotions based on autonomic nervous system indicators. The results implicated SDNN, RMSSD, CVRR, LF, and HF as important parameters for the predictions.

A Deep Learning Aided Drowning Diagnosis for Forensic Investigations using Post-Mortem Lung CT Images.

Feasibility of computer-aided diagnosis (CAD) systems has been demonstrated in the field of medical image diagnosis. Especially, deep learning based CAD systems showed high performance thanks to its capability of image recognition. However, there is no CAD system developed for post-mortem imaging diagnosis and thus it is still unclear if the CAD system is effective for this purpose. Particulally, the drowning diagnosis is one of the most difficult tasks in the field of forensic medicine because findings of the post-mortem image diagnosis are not specific. To address this issue, we develop a CAD system consisting of a deep convolution neural network (DCNN) to classify post-mortem lung computed tomography (CT) images into two categories of drowning and non-drowning cases. The DCNN was trained by means of transfer learning and performance evaluation was conducted by 10-fold cross validation using 140 drowning cases and 140 non-drowning cases of the CT images. The area under the receiver operating characteristic curve (AUC-ROC) for the DCNN was achieved 0.88 in average. This high performance clearly demonstrated that the proposed DCNN based CAD system has a potential for post-mortem image diagnosis of drowning.

Comparison of Visible and Infrared Video Plethysmography Captured from Different Regions of the Human Face.

Recently, video plethysmography (VPG) - a heart rate estimation technique using a video camera - has gained significant attention. Most studies of VPG have used a visible RGB camera; only a limited number of studies investigating near-infrared light (wavelength 750-2500 nm), which can be used even in a dark environment, have been performed. The purpose of this study was to investigate the differences between VPG data collected using visible light (VPGVIS) or near-infrared light (VPGNIR) from four facial areas (forehead, right cheek, left cheek, and nose). An experiment was conducted to obtain both VPGVIS and VPGNIR simultaneously by alternately irradiating the face with NIR and VIS lights. Experimental results showed that the root mean squared error of heart rate estimated using VPGNIR was 1 bpm higher than that of VPGVIS. However, contrary to our expectations, the power of the heartbeat-related component included in VPGNIR was not reduced despite the absorbance of hemoglobin in the NIR light range being 1/100 of that in the VIS light range. This result supports the hypothesis that a main factor in the generation of VPG waves was change in the optical properties caused by blood vessels compressing the subcutaneous tissue and the venous bed. Additionally, the accuracy of the heart rate estimation using VPG tended to be high when the nose was set as the ROI. This result was likely associated with the anatomical structure of the nose.

A method to measure slice sensitivity profiles of CT images under low-contrast and high-noise conditions.

Noise reduction features of iterative reconstruction (IR) methods in computed tomography might accompany the sacrifice of the longitudinal resolution, or slice sensitivity profile (SSP), at low contrast-to-noise ratio (CNR) conditions. To assess the benefit of IR methods correctly, the difference of SSP between IR methods and filtered-backprojection (FBP) must be taken into account. Therefore, SSP measurement under low-CNR conditions is necessary. Although edge methods are predominantly used, their performance under low-CNR conditions appears to be not fully established. We developed a method that is compatible with extremely low-CNR conditions. Thin plastic disk-shaped sheets embedded in acrylic resin were used as low-contrast test objects. The lowest peak contrast used was approximately 17 [HU]. We assessed the performance of our method by using FBP images. We identified a source of measurement instability aside from noise: the measured thin-slice SSP is dependent on the orbital phase of helical scan, presumably because of cone-beam artifacts. This impediment to high accuracy is manageable using phase-controlled scans. We confirmed that table position repeatability is much better than the value of the specifications, and therefore the ensemble-averaged images of multiple scans can be used for SSP measurement. Accurate measurement of SSP under extremely low-CNR conditions is possible, even when the test object is visually indiscernible from the noisy background. Low-contrast SSP behavior is elucidated for IR methods (AIDR-3D, FIRST, and AiSR-V) by using this measurement method.

Estimation of Absolute Blood Pressure Using Video Images Captured at Different Heights from the Heart.

The risk of cardiovascular diseases is related to the absolute level of blood pressure as well as its fluctuation while sleeping or during daily activities. To assess the risk, a simpler method to monitor daily blood pressure is desirable. In recent years, there has been a focus on developing a method to obtain pulse waves from video images of the human body. This is a promising technique to acquire biometric information without contact. In this study, we propose a new method to estimate the absolute level of blood pressure by using two video images of human hands captured at different heights from the heart. We focus on the amplitude difference of pulse waves obtained from the video images and derive an equation to estimate blood pressure based on the relationship between the internal pressure and the cross-sectional area of the blood vessel. The accuracy of the estimation was evaluated using data obtained from 5 healthy subjects performing cycling exercises that change their blood pressure. The average value of the root mean square error between the real value and the estimated value was 25.7 mmHg, while that of correlation coefficient was 0.66. There were large individual differences, particularly in the estimation of the absolute value of blood pressure. This result suggests the need for individual correction of the compliance curve, which represents the relationship between the internal pressure and the cross-sectional area of the blood vessel.

Learning Entropy as a Learning-Based Information Concept.

Recently, a novel concept of a non-probabilistic novelty detection measure, based on a multi-scale quantification of unusually large learning efforts of machine learning systems, was introduced as learning entropy (LE). The key finding with LE is that the learning effort of learning systems is quantifiable as a novelty measure for each individually observed data point of otherwise complex dynamic systems, while the model accuracy is not a necessary requirement for novelty detection. This brief paper extends the explanation of LE from the point of an informatics approach towards a cognitive (learning-based) information measure emphasizing the distinction from Shannon's concept of probabilistic information. Fundamental derivations of learning entropy and of its practical estimations are recalled and further extended. The potentials, limitations, and, thus, the current challenges of LE are discussed.

Potential improvements of lung and prostate MLC tracking investigated by treatment simulations.

PURPOSE/OBJECTIVES: Intrafraction tumor motion during external radiotherapy is a challenge for the treatment accuracy. A novel technique to mitigate the impact of tumor motion is real-time adaptation of the multileaf collimator (MLC) aperture to the motion, also known as MLC tracking. Although MLC tracking improves the dosimetric accuracy, there are still residual errors. Here, we investigate and rank the performance of five prediction algorithms and seven improvements of an MLC tracking system by extensive tracking treatment simulations. MATERIALS AND METHODS: An in-house-developed MLC tracking simulator that has been experimentally validated against an electromagnetic-guided MLC tracking system was used to test the prediction algorithms and tracking system improvements. The simulator requires a Dicom treatment plan and a motion trajectory as input and outputs all motion of the accelerator during MLC tracking treatment delivery. For lung tumors, MLC tracking treatments were simulated with a low and a high modulation VMAT plan using 99 patient-measured lung tumor trajectories. For prostate, tracking was also simulated with a low and a high modulation VMAT plan, but with 695 prostate trajectories. For each simulated treatment, the tracking error was quantified as the mean MLC exposure error, which is the sum of the overexposed area (irradiated area that should have been shielded according to the treatment plan) and the underexposed area (shielded area that should have been irradiated). First, MLC tracking was simulated with the current MLC tracking system without prediction, with perfect prediction (Perfect), and with the following five prediction algorithms: linear Kalman filter (Kalman), kernel density estimation (KDE), linear adaptive filtering (LAF), wavelet-based multiscale autoregression (wLMS), and time variant seasonal autoregression (TVSAR). Next, MLC tracking was simulated using the best prediction algorithm and seven different tracking system improvements: no localization signal latency (a), doubled maximum MLC leaf speed (b), halved MLC leaf width (c), use of Y backup jaws to track motion perpendicular to the MLC leaves (d), dynamic collimator rotation for alignment of the MLC leaves with the dominant target motion direction (e), improvements 4 and 5 combined (f), and all improvements combined (g). RESULTS: All results are presented as the mean residual MLC exposure error compared to no tracking. In the prediction study, the residual MLC exposure error was 47.0% (no prediction), 45.1% (Kalman), 43.8% (KDE), 43.7% (LAF), 42.1% (wLMS), 40.1% (TVSAR), and 36.5% (Perfect) for lung MLC tracking. For prostate MLC tracking, it was 66.0% (no prediction), 66.9% (Kalman), and 63.4% (Perfect). For lung with TVSAR prediction, the residual MLC exposure error for the seven tracking system improvements was 37.2%(1), 38.3%(2), 37.4%(3), 34.2%(4), 30.6%(5), 27.7%(6), and 20.7%(7). For prostate with no prediction, the residual MLC exposure error was 61.7%(1), 61.4%(2), 55.4%(3), 57.2%(4), 47.5%(5), 43.7%(6), and 38.7%(7). CONCLUSION: For prostate, MLC tracking was slightly better without prediction than with linear Kalman filter prediction. For lung, the TVSAR prediction algorithm performed best. Dynamic alignment of the collimator with the dominant motion axis was the most efficient MLC tracking improvement except for lung tracking with the low modulation VMAT plan, where jaw tracking was slightly better.