Dual Images Research Articles

A compressive sensing encryption scheme for dual color images based on discrete memristor map and Rubik’s cube scramble

The most common in practical image encryption are multiple color images, however, less attention is paid to inter-layer correlation broken in existing color image encryption algorithms. In this paper, a 2-D compressive sensing dual color images encryption algorithm based on a discrete memristor hyper-chaotic map and Rubik's cube scramble is proposed. First, we propose a new discrete memristor-Ikeda map (DM-Ikeda), the analysis result shows the DM-Ikeda map has extremely high complexity and hyper-chaotic dynamical behavior. Second, to facilitate ciphertext transmission, the images are compressed and encrypted simultaneously by DM-Ikeda map and 2D compressive sensing. Then, we design a new rubik's cube scramble algorithm, which combines the layers of two images into a cube that can be expanded in different orders, and swaps the pixels of rows, columns or faces of each image by rotating the cube to achieve image scramble. This algorithm can break the correlation not only of pixels within layers of two images but also of pixels between different layers of images. Finally, two images are further diffused with the use of hyper-chaotic sequences. The algorithm can achieve both compression and encryption of dual color images. Experimental results indicate that, our scheme has excellent compression, and can break the correlation between different images and layers, which is excellent resistance to statistical attacks and has high security.

Long-term Outcome of Anterior Chamber Parameters after Laser Iridotomy and Iridoplasty in Primary Angle Closure Glaucoma

Purpose: To investigate the long-term outcomes of anterior chamber angle parameters in patients with primary angle closure glaucoma (PACG) after peripheral laser iridotomy (LI) combined with iridoplasty (PI) using a dual Scheimpflug analyzer.Methods: This retrospective study included 32 eyes in 32 patients diagnosed with PACG who underwent LI plus PI. Patients with an acute angle closure crisis were excluded. Dual Scheimpflug analyzer images were acquired before and 1 week after LI plus PI, and yearly afterwards. Anterior chamber depth (ACD) and volume (ACV), mean anterior chamber angle (ACA), and intraocular pressure (IOP) were also measured.Results: The mean follow-up was 32.28 ± 13.34 months. Baseline demographics were age 63 ± 7.9 years, 62.5% female, IOP 15.48 ± 4.79 mmHg, ACD 2.09 ± 0.19 mm, and visual field mean deviation -7.97 ± 8.49 dB. ACD increased from baseline to 2.15 ± 0.32 mm, but it was not significant (<i>p</i> = 0.136). ACV increased significantly from 78.32 ± 11.49 mm at baseline to 83.04 ± 11.16 mm at the last visit after LI (<i>p</i> = 0.011). The mean ACA increased significantly from 26.86 ± 2.53° at baseline to 28.82 ± 4.64° at the last visit (<i>p</i> = 0.022). IOP decreased significantly from baseline to 13.06 ± 2.21 mmHg at the last visit (<i>p</i> = 0.001).Conclusions: The ACA parameters improved after LI combined with PI in patients with PACG and remained so on long-term follow-up. IOP was also significantly reduced for more than 2.5 years after LI plus PI.

Capacity-Raising Reversible Data Hiding Using Empirical Plus–Minus One in Dual Images

Electronic records of a patient’s health history are often shared among healthcare providers, and patient data must be kept secure to maintain the privacy of patients. One way of doing this is through data hiding, and this paper demonstrates a scheme to achieve this. This paper proposes a capacity-raising reversible data-hiding scheme using an empirical rules table in dual images. The aim of this research is to avoid drawing awareness to the transmission of information by providing a steganographic technique capable of embedding high-capacity data into an image while maintaining the good quality of the image. To hide the secret message(s), a rules table containing 13 entries is presented. This rules table is extendable to a table of up to 262,133 entries (with each entry containing one distinct character) that are related to the 13 entries in terms of the rules. The rules of this table are used during the embedding and extraction procedures. In the proposed method, 512 × 512 images are divided into 1 × 2 blocks where adjacent pixels are represented using x and y for both embedding and extraction, respectively. Recovery of the cover image from the stego image is also achievable during the extraction process. Conducted experiments show that the proposed method has an average pixel-to-signal noise ratio of 52.65 dB, which is higher than that achieved with the methods discussed in this paper. Additionally, the proposed method can embed a wider range of characters (depending on the image size) as compared to the rest of the methods, hence its high embedding capacity of 4.25 bpp. The proposed method can also withstand security attacks such as RS, pixel value difference, entropy, and chi-square attacks. The proposed method is also undetectable under visual attack analyses such as the difference histogram, pixel difference histogram, and visual inspection. Based on the higher embedding capacity, pixel-to-signal noise ratio, the ability of this method to be undetected under visual attack analysis, and the ability of this method to withstand security attacks, it can be concluded that the proposed method is superior to the other methods.

A high quality interpolation-based reversible data hiding technique using dual images

A high quality interpolation-based reversible data hiding technique using dual images

Integrated Small Animal PET/CT/RT with Onboard PET/CT Image Guidance for Preclinical Radiation Oncology Research.

We have integrated a compact and lightweight PET with an existing CT image-guided small animal irradiator to enable practical onboard PET/CT image-guided preclinical radiation therapy (RT) research. The PET with a stationary and full-ring detectors has ~1.1 mm uniform spatial resolution over its imaging field-of-view of 8.0 cm diameter and 3.5 cm axial length and was mechanically installed inside the irradiator in a tandem configuration with CT and radiation unit. A common animal bed was used for acquiring sequential dual functional and anatomical images with independent PET and CT control and acquisition systems. The reconstructed dual images were co-registered based on standard multi-modality image calibration and registration processes. Phantom studies were conducted to evaluate the integrated system and dual imaging performance. The measured mean PET/CT image registration error was ~0.3 mm. With one-bed and three-bed acquisitions, initial tumor focused and whole-body [18F]FDG animal images were acquired to test the capability of onboard PET/CT image guidance for preclinical RT research. Overall, the results have shown that integrated PET/CT/RT can provide advantageous and practical onboard PET/CT image to significantly enhance the accuracy of tumor delineation and radiation targeting that should enhance the existing and enable new and potentially breakthrough preclinical RT research and applications.

Bimodal CNN for cardiovascular disease classification by co-training ECG grayscale images and scalograms

This study aimed to develop a bimodal convolutional neural network (CNN) by co-training grayscale images and scalograms of ECG for cardiovascular disease classification. The bimodal CNN model was developed using a 12-lead ECG database collected from Chapman University and Shaoxing People's Hospital. The preprocessed database contains 10,588 ECG data and 11 heart rhythms labeled by a specialist physician. The preprocessed one-dimensional ECG signals were converted into two-dimensional grayscale images and scalograms, which are fed simultaneously to the bimodal CNN model as dual input images. The proposed model aims to improve the performance of CVDs classification by making use of ECG grayscale images and scalograms. The bimodal CNN model consists of two identical Inception-v3 backbone models, which were pre-trained on the ImageNet database. The proposed model was fine-tuned with 6780 dual-input images, validated with 1694 dual-input images, and tested on 2114 dual-input images. The bimodal CNN model using two identical Inception-v3 backbones achieved best AUC (0.992), accuracy (95.08%), sensitivity (0.942), precision (0.946) and F1-score (0.944) in lead II. Ensemble model of all leads obtained AUC (0.994), accuracy (95.74%), sensitivity (0.950), precision (0.953), and F1-score (0.952). The bimodal CNN model showed better diagnostic performance than logistic regression, XGBoost, LSTM, single CNN model training with grayscale images alone or with scalograms alone. The proposed bimodal CNN model would be of great help in diagnosing cardiovascular diseases.

An improved blind watermarking method facing dual color images based on Hadamard transform

An improved blind watermarking method facing dual color images based on Hadamard transform

Geometric Duality Results and Approximation Algorithms for Convex Vector Optimization Problems

We study geometric duality for convex vector optimization problems. For a primal problem with a -dimensional objective space, we formulate a dual problem with a -dimensional objective space. Consequently, different from an existing approach, the geometric dual problem does not depend on a fixed direction parameter, and the resulting dual image is a convex cone. We prove a one-to-one correspondence between certain faces of the primal and dual images. In addition, we show that a polyhedral approximation for one image gives rise to a polyhedral approximation for the other. Based on this, we propose a geometric dual algorithm which solves the primal and dual problems simultaneously and is free of direction-biasedness. We also modify an existing direction-free primal algorithm in such a way that it solves the dual problem as well. We test the performance of the algorithms for randomly generated problem instances by using the so-called primal error and hypervolume indicator as performance measures.

Semi-supervised interactive fusion network for MR image segmentation.

Medical image segmentation is an important task in the diagnosis and treatment of cancers. The low contrast and highly flexible anatomical structure make it challenging to accurately segment the organs or lesions. To improve the segmentation accuracy of the organs or lesions in magnetic resonance (MR) images, which can be useful in clinical diagnosis and treatment ofcancers. First, a selective feature interaction (SFI) module is designed to selectively extract the similar features of the sequence images based on the similarity interaction. Second, a multi-scale guided feature reconstruction (MGFR) module is designed to reconstruct low-level semantic features and focus on small targets and the edges of the pancreas. Third, to reduce manual annotation of large amounts of data, a semi-supervised training method is also proposed. Uncertainty estimation is used to further improve the segmentation accuracy. Three hundred ninety-five 3D MR images from 395 patients with pancreatic cancer, 259 3D MR images from 259 patients with brain tumors, and four-fold cross-validation strategy are used to evaluate the proposed method. Compared to state-of-the-art deep learning segmentation networks, the proposed method can achieve better segmentation of pancreas or tumors in MR images. SFI-Net can fuse dual sequence MR images for abnormal pancreas or tumor segmentation. The proposed semi-supervised strategy can further improve the performance ofSFI-Net.

Multi-Hider Reversible Data Hiding Using a Weighted Color Transfer and Modulus Operation

This paper proposes a multi-hider dual-image reversible data-hiding algorithm. Generating dual stego images, the proposed algorithm provides a high embedding capacity, exhibits high image quality, offers reversibility to restore the stego images, and suggests seven levels of secret message extraction. The message embedding in our algorithm contains two phases. In Phase-1, we embed an n-ary secret message vector followed by an m-ary secret message in each pixel using two distinct secret keys when rendering a new image from the source and target images using an optimal weighted color-transfer process. This phase produces a tentative stego image which exhibits a new color appearance, different from the input source image. In Phase-2, we introduce a weighted modulus operation to embed a k-ary secret message vector into a dual-pixel constructed from the tentative stego image. The message concealment is elaborately designed so that the hidden secret messages are intact. This phase produces dual stego images which carry three distinct secret messages. Using legitimate secret keys, an authorized receiver can extract one or parts of secret messages and recover the tentative stego color-transferred image without arousing any suspicion. The experimental results and analysis confirm that our scheme can resist RS and PVD steganalytic attacks. In addition, our algorithm provides both high embedding capacity and better image quality, outperforming its counterparts. To the best of our knowledge, our scheme is the first in the relevant literature that provides multi-hider reversible data hiding, thereby offering various levels of message extraction for secure data communication.

Improved Dual-image Quality with Reversible Data Hiding Using Translocation and Switching Strategy

Dual-image reversible data hiding (RDH) is a technique for hiding important messages. This technology can be used to safely deliver secret messages to the recipient through dual images in an open network without being easily noticed. The recipient of the image must receive the two stego-images before the secret message can be completely retrieved. Imperceptibility is one of the main advantages of data hiding technology; to increase the imperceptibility, the quality requirements of the stego-images are relatively important. A dual steganographic image RDH method, called a DS-CF scheme that can achieve a better steganographic image quality using the center folding (CF) strategy. In this paper, we developed a translocation and switching strategy (TaS) to shorten the distances between the stego-pixel coordinates and the cover pixel coordinates after information being hidden. Compared with the DS-CF scheme, our proposed DS-TaS scheme can effectively improve the quality of the steganographic images at the same level of embedding capability. The experimental results show that the PSNR of our DS-TaS scheme at = 1 was 55.66 dB, which is an increase of 1.5 dB, and is 51.43 dB for = 2, 46.66 dB for = 3, and 40.91 dB for = 4. In addition, the PSNR values of the stego images was increased by 1.5, 0.29, 0.29, and 0.19 dB, respectively. This shows that our proposed dual-image RDH method can optimize the visual quality of the stego-images and is better than many other dual-image RDH techniques.

Robust dual color images watermarking scheme with hyperchaotic encryption based on quaternion DFrAT and genetic algorithm

A robust dual color images watermarking algorithm is designed based on quaternion discrete fractional angular transform (QDFrAT) and genetic algorithm. To guarantee the watermark security, the original color watermark image is encrypted with a 4D hyperchaotic system. A pure quaternion matrix is acquired by performing the discrete wavelet transform (DWT), the block division and the discrete cosine transform on the original color cover image. The quaternion matrix is operated by the QDFrAT to improve the robustness and the security of the watermarking scheme with the optimal transform angle and the fractional order. Then the singular value matrix is obtained by the quaternion singular value decomposition (QSVD) to further enhance the scheme’s stability. The encryption watermark is also processed by DWT and QSVD. Afterward, the singular value matrix of the encryption watermark is embedded into the singular value matrix of the host image by the optimal scaling factor. Moreover, the values to balance imperceptibility and robustness are optimized with a genetic algorithm. It is shown that the proposed color image watermarking scheme performs well in imperceptibility, security, robustness and embedding capacity.

Artificially intelligent differential diagnosis of enlarged lymph nodes with random vector functional link network plus.

Differential diagnosis of enlarged lymph nodes (ELNs) is essential for the treatment of related patients. Though multi-modal ultrasound including B-mode, Doppler ultrasound, elastographyand contrast-enhanced ultrasound (CEUS)can enhance diagnostic performance for ELNs, the scenario of having only single or dual modal data is often encountered. In this study, an artificially intelligent diagnosis model based on the learning using privileged information was proposed to aid in differential diagnosis of ELNs in the case of single or dual modal images. In our model, B-mode, or combined with another modality, was used as the standard information (SI) and other modalities were used as the privileged information (PI). The model was constructed through the combination of the SI and PI in the training stage. By learning from the training samples, a random vector functional link network with privileged information (RVFL+) was obtained, which was used to classify the testing samples of solely the SI. Results showed that the accuracy, precision and Youden's index of the RVFL+ model, using B-mode with elastography as the SI and CEUS as the PI, reached 78.4%, 92.4% and 54.9%, increased by 14.0%, 8.4% and 24.5% compared with the model using B-mode as the SI without the PI. The method based on the LUPI can improve the diagnostic performance for ELNs.

Orthogonal integration of holographic and fluorescent dual images based on energy transfer from liquid crystals to a photocleavable AIEgen

Orthogonal integration of holographic and fluorescent dual-mode optical images with facile tunable emission colors has been achieved by precisely controlling the Förster resonance energy transfer and photocleavage reaction of an AIEgen.

Bifunctional Manipulation of Terahertz Waves with High‐Efficiency Transmissive Dielectric Metasurfaces

Multifunctional terahertz (THz) devices in transmission mode are highly desired in integration-optics applications, but conventional devices are bulky in size and inefficient. While ultra-thin multifunctional THz devices are recently demonstrated based on reflective metasurfaces, their transmissive counterparts suffer from severe limitations in efficiency and functionality. Here, based on high aspect-ratio silicon micropillars exhibiting wide transmission-phase tuning ranges with high transmission-amplitudes, a set of dielectric metasurfaces is designed and fabricated to achieve efficient spin-multiplexed wavefront controls on THz waves. As a benchmark test, the photonic-spin-Hall-effect is experimentally demonstrated with a record high absolute efficiency of 92% using a dielectric metasurface encoded with geometric phases only. Next, spin-multiplexed controls on circularly polarized THz beams (e.g., anomalous refraction and focusing) are experimentally demonstrated with experimental efficiency reaching 88%, based on a dielectric meta-device encoded with both spin-independent resonant phases and spin-dependent geometric phases. Finally, high-efficiency spin-multiplexed dual holographic images are experimentally realized with the third meta-device encoded with both resonant and geometric phases. Both near-field and far-field measurements are performed to characterize these devices, yielding results in agreement with full-wave simulations. The study paves the way to realize multifunctional, high-performance, and ultra-compact THz devices for applications in biology sensing, communications, and so on.

Simplification of dosimetry in 90Y-radioembolization therapy by dual planar images.

The purpose was to provide apractical and effective method for performing reliable 90Y dosimetry based on 99mTc-MAA and SPEC/CT. The impact of scatter correction (SC) and attenuation correction (AC) on the injected 90Y activity, lung shunt fraction (LSF) and the delivered dose to lung and liver compartments was investigated within the scope of the study. Eighteen eligible patients (F: 3, M: 15) were subjected to 90Y therapy. 99mTc-MAA (111-222 MBq) was injected into the targeted liver, followed by whole-body scan (WBS) with peak-window at 140 keV (15% width) and one down-scatter window. SPECT/CT scan was subsequently acquired encompassing lung and liver regions. The LSFs were fashioned from standard WBS LSFwb (St), scatter corrected WBS LSFwb (Sc), only scatter corrected SPECT LSFspect (NoAC-SC) and SPECT/CT with attenuation and scatter correction LSFspect (AC-SC). The absorbed doses that would be delivered to tumor and injected healthy liver were estimated using different calculation modes involving AC-SC (SPECT/CT), NoAC-SC (SPECT), NoAC-NoSC+LSFwb (SC), AC-SC + LSFwb (St), and NoAC-NoSC+LSFwb (St). The average deviations (range) in LSF values between standard LSFwb (St) and those from SPECT/CT (AC-SC), SPECT (NoAC-SC), and LSFwb (SC) were - 50% (- 29/- 71), - 32% (- 8/- 67), and - 45% (- 13/80), respectively. The suggested 90Y activity (GBq/Gy) was decreased within a range of 2-11%, 1-9%, and 2-7% by using LSFspect (AC-SC), LSFspect (NoAC-SC), and LSFwb (SC), respectively. Overall, two-sample t-test yielded no statistically significant difference (p < 0.05) in the absorbed doses to tumor and injected healthy liver between AC-SC (SPECT) and the rest of approaches with/and without AC and SC. However, a statistically significant difference (p < 0.05) was demonstrated in the lung shunt fractions and lung doses due to AC and SC. The LSFs from scatter corrected planar images LSFwb (SC) exhibited well agreement (R2 = 0.92) with SPECT/CT (AC-SC) and there was no statistically significant difference (Pvalue > 0.05) between both methods. It was deduced that SPECT/CT with attenuation and scatter correction plays a crucial role in the measurements of lung shunt fraction and dose as well as the total number of 90Y treatments. However, the absorbed dose to tumors and injected healthy liver was minimally affected by AC and SC. Besides, a good agreement was observed between LSF datasets from SPECT/CT versus scatter corrected WBS that can be alternatively and effectively used in 90Y dosimetry.

Snow Depth Inversion in Forest Areas from Sentinel-1 Data Based on Phase Deviation Correction

At present, snow depth inversion based on active microwave remote sensing is concerned essentially with areas having a relatively simple underlying surface. The existence of forests reduces the sensitivity of microwaves to snow, which often makes the snow depth inversion results uncertain. This paper presents a snow depth estimation algorithm for forest areas by introducing a forest phase to characterize the effect of forests on backscattering electromagnetic wave. Firstly, the interferogram is generated with the differential interference of two-pass master-slave Synthetic Aperture Radar (SAR) images, and the real phase under snow cover condition is obtained by phase unwrapping. Secondly, the phase models for forest and non-forest areas are constructed. The effects of forest cover are modeled as forest phase in the forest phase model, which is estimated under the assumption of snow depth consistency on both sides of the boundaries between forest and non-forest areas. Finally, snow depth is estimated by the snow phase-depth model. The correctness of the proposed forest snow depth inversion algorithm was verified by taking the Jiagedaqi area of Greater Xing’an Mountains as the study area and sentinel-1 dual polarization images as the data source. Finally, the snow depth distribution of the study area was obtained with a spatial resolution of 30 m on 7 December 2020. The experimental results show that the snow depth values estimated in Jiagedaqi area are mainly between 40–120 cm, and the average snow depth value is 80.27 cm. Taking the snow depth value of 84.69 cm reckoned from hourly accumulated snowfall in Jiagedaqi as the reference snow depth, the results of the estimated snow depth are relatively consistent and well-founded. With the introduction of the forest phase, the average snow depth values estimated in the forest area increase by 5.98 cm, which reduces the underestimation of the snow depth in forest areas.

How to Pseudo-CT: A Comparative Review of Deep Convolutional Neural Network Architectures for CT Synthesis

This paper provides an overview of the different deep convolutional neural network (DCNNs) architectures that have been investigated in the past years for the generation of synthetic computed tomography (CT) or pseudo-CT from magnetic resonance (MR). The U-net, the Atrous-net and the Residual-net architectures were analyzed, implemented and compared. Each network was implemented using 2D filters and 3D filters with 2D slices and 3D patches respectively as inputs. Two datasets were used for training and evaluation. The first one is composed by pairs of 3D T1-weighted MR and Low-dose CT images from the head of 19 healthy women. The second database contains dual echo Dixon-VIBE MR images and CT images from the pelvis of 13 colorectal and 6 prostate cancer patients. Bone structures in the target anatomy were key in choosing the right deep learning approach. This work provides a deep explanation of the architectures in order to know which DCNN fits better each medical application. According to this study, the 3D U-net architecture would be the best option to generate head pseudo-CTs while the 2D Residual-net provides the most accurate results for the pelvis anatomy.

A New Dual Image Based Reversible Data Hiding Method Using Most Significant Bits and Center Shifting Technique

In this article, a new reversible data hiding method using most significant bits and center shifting technique in dual images is proposed. The proposed reversible data hiding method aims to securely hide high-capacity secret data. Instead of directly embedding the secret data, the method calculates new values with (n + 1) bits secret data and the n most significant bits of the cover image pixel values. Thus, it is impossible to extract secret data without obtaining the original cover image. Also, the center shifting process is performed to minimize the mean square error. After this process, the secret data to be hidden in the dual images are in the range [−3×2(n−1)+1, 3×2(n−1)−1]. The pixel values of stego images are obtained by using the secret data in this range value and cover image pixel values. As a result of experimental studies, when the payload is 2.5 bits per pixel (bpp), the peak signal-to-noise ratio value (PSNR), which expresses the visual quality, is above 34 (dB). In addition, the proposed method has proven secure against RS (regular and singular) analysis attacks.

Evaluation of Markerless Tumor Tracking Using Synthetic Dual Energy Images Produced Using a Convolutional Neural Network

<h3>Purpose/Objective(s)</h3> Dual-energy (DE) fluoroscopy is being considered for markerless tumor tracking (MTT) of lung tumors. The principal advantage of DE imaging is that it removes overlying bony anatomy and improves tumor tracking accuracy, compared with single energy (SE) fluoroscopy. However, to be used clinically, DE imaging requires specific hardware/software which is not widely available. To address this limitation, we developed and implemented a convolutional neural network (CNN) to generate synthetic DE (sDE) images from SE images. The goal of this study is to evaluate the accuracy of MTT using sDE as compared to SE and real DE (rDE) images. <h3>Materials/Methods</h3> To evaluate MTT with DE fluoroscopy, a thoracic motion phantom, consisting of a torso with embedded ribs and spine along with a lung-type cavity, was used. A simulated tumor (10 mm diameter) was placed inside the lung equivalent compartment of the phantom and then programmed to simulate breathing (peak-to-peak amplitude=15 mm, period = 5 s) using a programmed cos⁴ function. While the target was moving, fast kV switching images (alternating 60 and 120 kVp) were obtained using the on-board imager (OBI) of a commercial linear accelerator. Images were acquired over 360 degrees of rotation, allowing the target projection to overlap with varying amounts of bone. Weighted logarithmic subtraction was performed offline on consecutive 60 and 120 kVp projections to produce DE images. Separately, a CNN (U-net) was trained to produce sDE directly from the 120 kVp images (which are used clinically for SE fluoroscopy). The CNN uses a "U-Net" type architecture, with skip connections used to pass low-level information across the network for use in image translation. A template-based matching algorithm was then used to track target motion on SE (120 kVp), rDE, and sDE images. A quantitative analysis was performed based on tracking success rate (TSR) which was evaluated against ground truth (GT) using the programmed cos⁴ waveform. <h3>Results</h3> A total of 449 SE, rDE and sDE image frames were evaluated for the thoracic phantom with a 10 mm moving target. TSR values based on < 1 mm agreement with GT on SE, rDE, and sDE images were 53.5%, 89.7%, and 74.2%, respectively. The percentage of images with a TSR within < 2 mm agreement with GT was 71.9% for SE, 93.5% for rDE and 89.5% for sDE. Lower TSR percentages were observed on sDE and SE images where the simulated tumor overlapped the spine. <h3>Conclusion</h3> sDE images generated using a CNN have a TSR that is higher than SE images, and have TSR values that are close to rDE images. Further development in the CNN is required to improve the tracking accuracy of sDE images in the spine region. The preliminary findings of this study suggest that the use of sDE images generated from SE images potentially allow for tracking with DE-like images on any OBI, without the need for additional hardware/software.