Deep learning-based coronary computed tomography analysis to predict functionally significant coronary artery stenosis.

Fractional flow reserve derived from coronary CT (FFR-CT) is a noninvasive physiological technique that has shown a good correlation with invasive FFR. However, the use of FFR-CT is restricted by strict application standards, and the diagnostic accuracy of FFR-CT analysis may potentially be decreased by severely calcified coronary arteries because of blooming and beam hardening artifacts. The aim of this study was to evaluate the utility of deep learning (DL)-based coronary computed tomography (CT) data analysis in predicting invasive fractional flow reserve (FFR), especially in cases with severely calcified coronary arteries. We analyzed 184 consecutive cases (241 coronary arteries) which underwent coronary CT and invasive coronary angiography, including invasive FFR, within a three-month period. Mean coronary artery calcium scores were 963 ± 1226. We evaluated and compared the vessel-based diagnostic accuracy of our proposed DL model and a visual assessment to evaluate functionally significant coronary artery stenosis (invasive FFR < 0.80). A deep neural network was trained with consecutive short axial images of coronary arteries on coronary CT. Ninety-one coronary arteries of 89 cases (48%) had FFR-positive functionally significant stenosis. On receiver operating characteristics (ROC) analysis to predict FFR-positive stenosis using the trained DL model, average area under the curve (AUC) of the ROC curve was 0.756, which was superior to the AUC of visual assessment of significant (≥ 70%) coronary artery stenosis on CT (0.574, P = 0.011). The sensitivity, specificity, positive and negative predictive value (PPV and NPV), and accuracy of the DL model and visual assessment for detecting FFR-positive stenosis were 82 and 36%, 68 and 78%, 59 and 48%, 87 and 69%, and 73 and 63%, respectively. Sensitivity and NPV for the prediction of FFR-positive stenosis were significantly higher with our DL model than visual assessment (P = 0.0004, and P = 0.024). DL-based coronary CT data analysis has a higher diagnostic accuracy for functionally significant coronary artery stenosis than visual assessment.

Improving respiratory signal prediction with a deep neural network and simple changes to the input and output data format.

Objective.To improve respiratory gating accuracy and radiation treatment throughput, we developed a generalized model based on a deep neural network (DNN) for predicting any given patient's respiratory motion.Approach.Our model uses long short-term memory (LSTM) based on a recurrent neural network (RNN), and improves upon common techniques. The first improvement is that the data input is not a one-dimensional sequence, but two-dimensional block data. This shortens the input sequence length, reducing computation time. Second, the output is not a scalar, but a sequence prediction. This increases the amount of available data, allowing improved prediction accuracy. For training and evaluation of our model, 434 sets of real-time position management data were retrospectively collected from clinical studies. The data were separated in a ratio of 4:1, with the larger set used for training models and the remaining set used for testing. We measured the accuracy of respiratory signal prediction and amplitude-based gating with prediction windows equaling 133, 333, and 533 ms. This new model was compared with the original LSTM and a non-recurrent DNN model.Main results.The mean absolute errors with the prediction window at 133, 333 and 533 ms were 0.036, 0.084, 0.119 with our model; 0.049, 0.14, 0.246 with the original LSTM-based model; and 0.041, 0.119, 0.16 with the non-recurrent DNN model, respectively. The computation time were 0.66 ms with our model; 0.63 ms the original LSTM-based model; 1.60 ms the non-recurrent DNN model, respectively. The accuracies of amplitude-based gating with the same prediction window settings and a duty cycle of approximately 50% were 98.3%, 95.8% and 92.7% with our model, 97.6%, 93.9% and 87.2% with the original LSTM-based model; and 97.9%, 94.3% and 89.5% with the non-recurrent DNN model, respectively.Significance.Our RNN algorithm for respiratory signal prediction successfully estimated tumor positions. We believe it will be useful in respiratory signal prediction technology.

Spatiotemporal forecasting of vertical track alignment with exogenous factors.

To ensure the safety of railroad operations, it is important to monitor and forecast track geometry irregularities. A higher safety requires forecasting with higher spatiotemporal frequencies, which in turn requires capturing spatial correlations. Additionally, track geometry irregularities are influenced by multiple exogenous factors. In this study, a method is proposed to forecast one type of track geometry irregularity, vertical alignment, by incorporating spatial and exogenous factor calculations. The proposed method embeds exogenous factors and captures spatiotemporal correlations using a convolutional long short-term memory. The proposed method is also experimentally compared with other methods in terms of the forecasting performance. Additionally, an ablation study on exogenous factors is conducted to examine their individual contributions to the forecasting performance. The results reveal that spatial calculations and maintenance record data improve the forecasting of vertical alignment.

Optimizing 3DCT image registration for interfractional changes in carbon-ion prostate radiotherapy.

To perform setup procedures including both positional and dosimetric information, we developed a CT-CT rigid image registration algorithm utilizing water equivalent pathlength (WEPL)-based image registration and compared the resulting dose distribution with those of two other algorithms, intensity-based image registration and target-based image registration, in prostate cancer radiotherapy using the carbon-ion pencil beam scanning technique. We used the data of the carbon ion therapy planning CT and the four-weekly treatment CTs of 19 prostate cancer cases. Three CT-CT registration algorithms were used to register the treatment CTs to the planning CT. Intensity-based image registration uses CT voxel intensity information. Target-based image registration uses target position on the treatment CTs to register it to that on the planning CT. WEPL-based image registration registers the treatment CTs to the planning CT using WEPL values. Initial dose distributions were calculated using the planning CT with the lateral beam angles. The treatment plan parameters were optimized to administer the prescribed dose to the PTV on the planning CT. Weekly dose distributions using the three different algorithms were calculated by applying the treatment plan parameters to the weekly CT data. Dosimetry, including the dose received by 95% of the clinical target volume (CTV-D95), rectal volumes receiving > 20 Gy (RBE) (V20), > 30 Gy (RBE) (V30), and > 40 Gy (RBE) (V40), were calculated. Statistical significance was assessed using the Wilcoxon signed-rank test. Interfractional CTV displacement over all patients was 6.0 ± 2.7 mm (19.3 mm maximum standard amount). WEPL differences between the planning CT and the treatment CT were 1.2 ± 0.6 mm-H2O (< 3.9 mm-H2O), 1.7 ± 0.9 mm-H2O (< 5.7 mm-H2O) and 1.5 ± 0.7 mm-H2O (< 3.6 mm-H2O maxima) with the intensity-based image registration, target-based image registration, and WEPL-based image registration, respectively. For CTV coverage, the D95 values on the planning CT were > 95% of the prescribed dose in all cases. The mean CTV-D95 values were 95.8 ± 11.5% and 98.8 ± 1.7% with the intensity-based image registration and target-based image registration, respectively. The WEPL-based image registration was CTV-D95 to 99.0 ± 0.4% and rectal Dmax to 51.9 ± 1.9 Gy (RBE) compared to 49.4 ± 9.1 Gy (RBE) with intensity-based image registration and 52.2 ± 1.8 Gy (RBE) with target-based image registration. The WEPL-based image registration algorithm improved the target coverage from the other algorithms and reduced rectal dose from the target-based image registration, even though the magnitude of the interfractional variation was increased.

Real-time deep neural network-based automatic bowel gas segmentation on X-ray images for particle beam treatment.

Since particle beam distribution is vulnerable to change in bowel gas because of its low density, we developed a deep neural network (DNN) for bowel gas segmentation on X-ray images. We used 6688 image datasets from 209 cases as training data, 736 image datasets from 23 cases as validation data and 102 image datasets from 51 cases as test data (total 283 cases). For the training data, we prepared three types of digitally reconstructed radiographic (DRR) images (all-density, bone and gas) by projecting the treatment planning CT image data. However, the real X-ray images acquired in the treatment room showed low contrast that interfered with manual delineation of bowel gas. Therefore, we used synthetic X-ray images converted from DRR images in addition to real X-ray images.We evaluated DNN segmentation accuracy for the synthetic X-ray images using Intersection over Union, recall, precision, and the Dice coefficient, which measured 0.708 ± 0.208, 0.832 ± 0.170, 0.799 ± 0.191, and 0.807 ± 0.178, respectively. The evaluation metrics for the real X-images were less accurate than those for the synthetic X-ray images (0.408 ± 0237, 0.685 ± 0.326, 0.490 ± 0272, and 0.534 ± 0.271, respectively). Computation time was 29.7 ± 1.3 ms/image. Our DNN appears useful in increasing treatment accuracy in particle beam therapy.

Deep learning-based gene selection in comprehensive gene analysis in pancreatic cancer.

The selection of genes that are important for obtaining gene expression data is challenging. Here, we developed a deep learning-based feature selection method suitable for gene selection. Our novel deep learning model includes an additional feature-selection layer. After model training, the units in this layer with high weights correspond to the genes that worked effectively in the processing of the networks. Cancer tissue samples and adjacent normal pancreatic tissue samples were collected from 13 patients with pancreatic ductal adenocarcinoma during surgery and subsequently frozen. After processing, gene expression data were extracted from the specimens using RNA sequencing. Task 1 for the model training was to discriminate between cancerous and normal pancreatic tissue in six patients. Task 2 was to discriminate between patients with pancreatic cancer (n = 13) who survived for more than one year after surgery. The most frequently selected genes were ACACB, ADAMTS6, NCAM1, and CADPS in Task 1, and CD1D, PLA2G16, DACH1, and SOWAHA in Task 2. According to The Cancer Genome Atlas dataset, these genes are all prognostic factors for pancreatic cancer. Thus, the feasibility of using our deep learning-based method for the selection of genes associated with pancreatic cancer development and prognosis was confirmed.

Development of attenuation correction methods using deep learning in brain-perfusion single-photon emission computed tomography.

PURPOSE: Computed tomography (CT)-based attenuation correction (CTAC) in single-photon emission computed tomography (SPECT) is highly accurate, but it requires hybrid SPECT/CT instruments and additional radiation exposure. To obtain attenuation correction (AC) without the need for additional CT images, a deep learning method was used to generate pseudo-CT images has previously been reported, but it is limited because of cross-modality transformation, resulting in misalignment and modality-specific artifacts. This study aimed to develop a deep learning-based approach using non-attenuation-corrected (NAC) images and CTAC-based images for training to yield AC images in brain-perfusion SPECT. This study also investigated whether the proposed approach is superior to conventional Chang's AC (ChangAC). METHODS: In total, 236 patients who underwent brain-perfusion SPECT were randomly divided into two groups: the training group (189 patients; 80%) and the test group (47 patients; 20%). Two models were constructed using Autoencoder (AutoencoderAC) and U-Net (U-NetAC), respectively. ChangAC, AutoencoderAC, and U-NetAC approaches were compared with CTAC using qualitative analysis (visual evaluation) and quantitative analysis (normalized mean squared error [NMSE] and the percentage error in each brain region). Statistical analyses were performed using the Wilcoxon signed-rank sum test and Bland-Altman analysis. RESULTS: U-NetAC had the highest visual evaluation score. The NMSE results for the U-NetAC were the lowest, followed by AutoencoderAC and ChangAC (P < 0.001). Bland-Altman analysis showed a fixed bias for ChangAC and AutoencoderAC and a proportional bias for ChangAC. ChangAC underestimated counts by 30-40% in all brain regions. AutoencoderAC and U-NetAC produced mean errors of <1% and maximum errors of 3%, respectively. CONCLUSION: New deep learning-based AC methods for AutoencoderAC and U-NetAC were developed. Their accuracy was higher than that obtained by ChangAC. U-NetAC exhibited higher qualitative and quantitative accuracy than AutoencoderAC. We generated highly accurate AC images directly from NAC images without the need for intermediate pseudo-CT images. To verify our models' generalizability, external validation is required.

Multidimensional Deep Learning Reduces False-Positives in the Automated Detection of Cerebral Aneurysms on Time-Of-Flight Magnetic Resonance Angiography: A Multi-Center Study.

Current deep learning-based cerebral aneurysm detection demonstrates high sensitivity, but produces numerous false-positives (FPs), which hampers clinical application of automated detection systems for time-of-flight magnetic resonance angiography. To reduce FPs while maintaining high sensitivity, we developed a multidimensional convolutional neural network (MD-CNN) designed to unite planar and stereoscopic information about aneurysms. This retrospective study enrolled time-of-flight magnetic resonance angiography images of cerebral aneurysms from three institutions from June 2006 to April 2019. In the internal test, 80% of the entire data set was used for model training and 20% for the test, while for the external tests, data from different pairs of the three institutions were used for training and the remaining one for testing. Images containing aneurysms > 15 mm and images without aneurysms were excluded. Three deep learning models [planar information-only (2D-CNN), stereoscopic information-only (3D-CNN), and multidimensional information (MD-CNN)] were trained to classify whether the voxels contained aneurysms, and they were evaluated on each test. The performance of each model was assessed using free-response operating characteristic curves. In total, 732 aneurysms (5.9 ± 2.5 mm) of 559 cases (327, 120, and 112 from institutes A, B, and C; 469 and 263 for 1.5T and 3.0T MRI) were included in this study. In the internal test, the highest sensitivities were 80.4, 87.4, and 82.5%, and the FPs were 6.1, 7.1, and 5.0 FPs/case at a fixed sensitivity of 80% for the 2D-CNN, 3D-CNN, and MD-CNN, respectively. In the external test, the highest sensitivities were 82.1, 86.5, and 89.1%, and 5.9, 7.4, and 4.2 FPs/cases for them, respectively. MD-CNN was a new approach to maintain sensitivity and reduce the FPs simultaneously.

A Deep Convolutional Neural Network With Performance Comparable to Radiologists for Differentiating Between Spinal Schwannoma and Meningioma.

STUDY DESIGN: Retrospective analysis of magnetic resonance imaging (MRI). OBJECTIVE: The aim of this study was to evaluate the performance of our convolutional neural network (CNN) in differentiating between spinal schwannoma and meningioma on MRI. We compared the performance of the CNN and that of two expert radiologists. SUMMARY OF BACKGROUND DATA: Preoperative discrimination between spinal schwannomas and meningiomas is crucial because different surgical procedures are required for their treatment. A deep-learning approach based on CNNs is gaining interest in the medical imaging field. METHODS: We retrospectively reviewed data from patients with spinal schwannoma and meningioma who had undergone MRI and tumor resection. There were 50 patients with schwannoma and 34 patients with meningioma. Sagittal T2-weighted magnetic resonance imaging (T2WI) and sagittal contrast-enhanced T1-weighted magnetic resonance imaging (T1WI) were used for the CNN training and validation. The deep learning framework Tensorflow was used to construct the CNN architecture. To evaluate the performance of the CNN, we plotted the receiver-operating characteristic (ROC) curve and calculated the area under the curve (AUC). We calculated and compared the sensitivity, specificity, and accuracy of the diagnosis by the CNN and two board-certified radiologists. RESULTS: . The AUC of ROC curves of the CNN based on T2WI and contrast-enhanced T1WI were 0.876 and 0.870, respectively. The sensitivity of the CNN based on T2WI was 78%; 100% for radiologist 1; and 95% for radiologist 2. The specificity was 82%, 26%, and 42%, respectively. The accuracy was 80%, 69%, and 73%, respectively. By contrast, the sensitivity of the CNN based on contrast-enhanced T1WI was 85%; 100% for radiologist 1; and 96% for radiologist 2. The specificity was 75%, 56, and 58%, respectively. The accuracy was 81%, 82%, and 81%, respectively. CONCLUSION: We have successfully differentiated spinal schwannomas and meningiomas using the CNN with high diagnostic accuracy comparable to that of experienced radiologists. LEVEL OF EVIDENCE: 4.

Nonextensive entropies derived from form invariance of pseudoadditivity.

The form invariance of pseudoadditivity is shown to determine the structure of nonextensive entropies. Nonextensive entropy is defined as the appropriate expectation value of nonextensive information content, similar to the definition of Shannon entropy. Information content in a nonextensive system is obtained uniquely from generalized axioms by replacing the usual additivity with pseudoadditivity. The satisfaction of the form invariance of the pseudoadditivity of nonextensive entropy and its information content is found to require the normalization of nonextensive entropies. The proposed principle requires the same normalization as that derived previously [A.K. Rajagopal and S. Abe, Phys. Rev. Lett. 83, 1711 (1999)], but is simpler and establishes a basis for the systematic definition of various entropies in nonextensive systems.