Classification Prediction of Alzheimer's Disease and Vascular Dementia Using Physiological Data and ECD SPECT Images.

Alzheimer's disease (AD) and vascular dementia (VaD) are the two most common forms of dementia. However, their neuropsychological and pathological features often overlap, making it difficult to distinguish between AD and VaD. In addition to clinical consultation and laboratory examinations, clinical dementia diagnosis in Taiwan will also include Tc-99m-ECD SPECT imaging examination. Through machine learning and deep learning technology, we explored the feasibility of using the above clinical practice data to distinguish AD and VaD. We used the physiological data (33 features) and Tc-99m-ECD SPECT images of 112 AD patients and 85 VaD patients in the Taiwanese Nuclear Medicine Brain Image Database to train the classification model. The results, after filtering by the number of SVM RFE 5-fold features, show that the average accuracy of physiological data in distinguishing AD/VaD is 81.22% and the AUC is 0.836; the average accuracy of training images using the Inception V3 model is 85% and the AUC is 0.95. Finally, Grad-CAM heatmap was used to visualize the areas of concern of the model and compared with the SPM analysis method to further understand the differences. This research method can quickly use machine learning and deep learning models to automatically extract image features based on a small amount of general clinical data to objectively distinguish AD and VaD.

The Feasibility of Differentiating Lewy Body Dementia and Alzheimer's Disease by Deep Learning Using ECD SPECT Images.

The correct differential diagnosis of dementia has an important impact on patient treatment and follow-up care strategies. Tc-99m-ECD SPECT imaging, which is low cost and accessible in general clinics, is used to identify the two common types of dementia, Alzheimer's disease (AD) and Lewy body dementia (LBD). Two-stage transfer learning technology and reducing model complexity based on the ResNet-50 model were performed using the ImageNet data set and ADNI database. To improve training accuracy, the three-dimensional image was reorganized into three sets of two-dimensional images for data augmentation and ensemble learning, then the performance of various deep learning models for Tc-99m-ECD SPECT images to distinguish AD/normal cognition (NC), LBD/NC, and AD/LBD were investigated. In the AD/NC, LBD/NC, and AD/LBD tasks, the AUC values were around 0.94, 0.95, and 0.74, regardless of training models, with an accuracy of 90%, 87%, and 71%, and F1 scores of 89%, 86%, and 76% in the best cases. The use of transfer learning and a modified model resulted in better prediction results, increasing the accuracy by 32% for AD/NC. The proposed method is practical and could rapidly utilize a deep learning model to automatically extract image features based on a small number of SPECT brain perfusion images in general clinics to objectively distinguish AD and LBD.

Detection of Alzheimer's disease using ECD SPECT images by transfer learning from FDG PET.

OBJECTIVE: To develop a practical method to rapidly utilize a deep learning model to automatically extract image features based on a small number of SPECT brain perfusion images in general clinics to objectively evaluate Alzheimer's disease (AD). METHODS: For the properties of low cost and convenient access in general clinics, Tc-99-ECD SPECT imaging data in brain perfusion detection was used in this study for AD detection. Two-stage transfer learning based on the Inception v3 network model was performed using the ImageNet dataset and ADNI database. To improve training accuracy, the three-dimensional image was reorganized into three sets of two-dimensional images for data augmentation and ensemble learning. The effect of pre-training parameters for Tc-99m-ECD SPECT image to distinguish AD from normal cognition (NC) was investigated, as well as the effect of the sample size of F-18-FDG PET images used in pre-training. The same model was also fine-tuned for the prediction of the MMSE score from the Tc-99m-ECD SPECT image. RESULTS: The AUC values of w/wo pre-training parameters for Tc-99m-ECD SPECT image to distinguish AD from NC were 0.86 and 0.90. The sensitivity, specificity, precision, accuracy, and F1 score were 100%, 75%, 76%, 86%, and 86%, respectively for the training model with 1000 cases of F-18-FDG PET image for pre-training. The AUC values for various sample sizes of the training dataset (100, 200, 400, 800, 1000 cases) for pre-training were 0.86, 0.91, 0.95, 0.97, and 0.97. Regardless of the pre-training condition ECD dataset used, the AUC value was greater than 0.85. Finally, predicting cognitive scores and MMSE scores correlated (R2 = 0.7072). CONCLUSIONS: With the ADNI pre-trained model, the sensitivity and accuracy of the proposed deep learning model using SPECT ECD perfusion images to differentiate AD from NC were increased by approximately 30% and 10%, respectively. Our study indicated that the model trained on PET FDG metabolic imaging for the same disease could be transferred to a small sample of SPECT cerebral perfusion images. This model will contribute to the practicality of SPECT cerebral perfusion images using deep learning technology to objectively recognize AD.

Calculation of effective atomic numbers using a rational polynomial approximation method with a dual-energy X-ray imaging system.

The study aims to develop a rational polynomial approximation method for improving the accuracy of the effective atomic number calculation with a dual-energy X-ray imaging system. This method is based on a multi-materials calibration model with iterative optimization, which can improve the calculation accuracy of the effective atomic number by adding a rational term without increasing the computation time. The performance of the proposed rational polynomial approximation method is demonstrated and validated by both simulated and experimental studies. The twelve reference materials are used to establish the effective atomic number calibration model, and the value of the effective atomic numbers are between 5.444 and 22. For the accuracy of the effective atomic number calculation, the relative differences between calculated and experimental values are less than 8.5%for all sample cases in this study. The average calculation accuracy of the method proposed in this study can be improved by about 40%compared with the conventional polynomial approximation method. Additionally, experimental quality assurance phantom imaging result indicates that the proposed method is compliant with the international baggage inspection standards for detecting the explosives. Moreover, the experimental imaging results reveal that the difference of color between explosives and the surrounding materials is in significant contrast for the dual-energy image with the proposed method.

Time of flight dual photon emission computed tomography.

Time-of-flight dual photon emission computed tomography (TOF-DuPECT) is an imaging system that can obtain radionuclide distributions using time information recorded from two cascade-decay photons. The potential decay locations in the image space, a hyperbolic response curve, can be determined via time-difference-of-arrival (TDOA) estimations from two instantaneous coincidence photons. In this feasibility study, Monte Carlo simulations were performed to generate list-mode coincidence data. A full-ring positron emission tomography-like detection system geometry was built in the simulation environment. A contrast phantom and a Jaszczak-like phantom filled with Selenium-75 (Se-75) were used to evaluate the image quality. A TOF-DuPECT system with varying coincidence time resolution (CTR) was then evaluated. We used the stochastic origin ensemble (SOE) algorithm to reconstruct images from the recorded list-mode data. The results indicate that the SOE method can be successfully employed for the TOF-DuPECT system and can achieve acceptable image quality when the CTR is less than 100 ps. Therefore, the TOF-DuPECT imaging system is feasible. With the improvement of the detector with time, future implementations and applications of TOF-DuPECT are promising. Further quantitative imaging techniques such as attenuation and scatter corrections for the TOF-DuPECT system will be developed in future.

A noise smoothing origin ensemble algorithm based on local filtering.

An origin ensemble (OE) image reconstruction algorithm can be used for the fast reconstruction of unconventional geometrical images, e.g. in a Compton camera (CC) system. Due to the low-count rate in the emission data, the reconstructed image is often noisy and inhomogeneous in density. In this study, we propose a way to smooth out the noise in the OE algorithm. During the OE reconstruction, the algorithm stochastically modifies the current location to a random new voxel along the probable corresponding curve of each event depending on the relative event density of the new and old locations. In the original OE technique, the event density is simply the number of events in the voxel. In the proposed method, the event density is estimated from the filtering of a kernel window centered on the voxel. Incorporating the regional filtering is similar to performing an OE algorithm on a smoothed image at each iteration and enables the reconstruction of a smoother image. A Flangeless Esser PET phantom and a multi-activity phantom are used to study the property of the new reconstruction algorithm. The results indicate that the proposed method performs better than a conventional OE algorithm in terms of normalized mean square error (NMSE) and structural similarity (SSIM). Both contrast noise ratio (CNR) and reconstruction accuracy of the new method are better than the conventional OE algorithm and their performances improve with the increase of object size. The median-OE possesses the highest overall image quality and recovery rate among the three filter-OE algorithms and is the method of choice for image reconstruction. Comparing to conventional post-smoothing OEs, the NMSE of median-OE improves 57.6% (46.9%) and the SSIM increased by 73.2% (51.1%) for the Esser (multi-activity) phantom. The proposed OE algorithm is simple and efficient for noise smoothing without complex calculations and highly suited for low-count cases.

A geometric calibration method for the digital chest tomosynthesis with dual-axis scanning geometry.

The aim of this study was to develop a geometric calibration method capable of eliminating the reconstruction artifacts of geometric misalignments in a tomosynthesis prototype with dual-axis scanning geometry. The potential scenarios of geometric misalignments were demonstrated, and their effects on reconstructed images were also evaluated. This method was a phantom-based approach with iterative optimization, and the calibration phantom was designed for specific tomosynthesis scanning geometry. The phantom was used to calculate a set of geometric parameters from each projection, and these parameters were then used to evaluate the geometric misalignments of the dual-axis scanning-geometry prototype. The simulated results revealed that the extracted geometric parameters were similar to the input values and that the artifacts of reconstructed images were minimized due to geometric calibration. Additionally, experimental chest phantom imaging results also indicated that the artifacts of the reconstructed images were suppressed and that object structures were preserved through calibration. And the quantitative analysis result also indicated that the MTF can be further improved with the geometric calibration. All the simulated and experimental results demonstrated that this method is effective for tomosynthesis with dual-axis scanning geometry. Furthermore, this geometric calibration method can also be applied to other tomography imaging systems to reduce geometric misalignments and be used for different geometric calibration phantom configurations.

Noninvasive measurement of radiopharmaceutical time-activity data using external thermoluminescent dosimeters (TLDs).

In this study, we present a new method for estimating the time-activity data using serial timely measurements of thermoluminescent dosimeters (TLDs). The approach is based on the combination of the measurement of surface dose using TLD and Monte Carlo (MC) simulation to estimate the radiopharmaceutical time-activity data. It involves four steps: (1) identify the source organs and outline their contours in computed tomography images; (2) compute the S values on the body surface for each source organ using a MC code; (3) obtain a serial measurement of the dose with numerous TLDs placed on the body surface; (4) solve the dose-activity equation to generate organ cumulative activity for each period of measurement. The activity of each organ at the time of measurement is simply the cumulative activity divided by the timespan between measurements. The usefulness of this method was studied using a MC simulation based on an Oak Ridge National Laboratory mathematical phantom with 18F-FDG filled in six source organs. Numerous TLDs were placed on different locations of the surface and were repeatedly read and replaced. The time-activity curves (TACs) of all organs were successfully reconstructed. Experiments on a physical phantom were also performed. Preliminary results indicate that it is an effective, robust, and simple method for assessing the TAC. The proposed method holds great potential for a range of applications in areas such as targeted radionuclide therapy, pharmaceutical research, and patient-specific dose estimation.

Efficient simulation of voxelized phantom in GATE with embedded SimSET multiple photon history generator.

GEANT4 Application for Tomographic Emission (GATE) is a powerful Monte Carlo simulator that combines the advantages of the general-purpose GEANT4 simulation code and the specific software tool implementations dedicated to emission tomography. However, the detailed physical modelling of GEANT4 is highly computationally demanding, especially when tracking particles through voxelized phantoms. To circumvent the relatively slow simulation of voxelized phantoms in GATE, another efficient Monte Carlo code can be used to simulate photon interactions and transport inside a voxelized phantom. The simulation system for emission tomography (SimSET), a dedicated Monte Carlo code for PET/SPECT systems, is well-known for its efficiency in simulation of voxel-based objects. An efficient Monte Carlo workflow integrating GATE and SimSET for simulating pinhole SPECT has been proposed to improve voxelized phantom simulation. Although the workflow achieves a desirable increase in speed, it sacrifices the ability to simulate decaying radioactive sources such as non-pure positron emitters or multiple emission isotopes with complex decay schemes and lacks the modelling of time-dependent processes due to the inherent limitations of the SimSET photon history generator (PHG). Moreover, a large volume of disk storage is needed to store the huge temporal photon history file produced by SimSET that must be transported to GATE. In this work, we developed a multiple photon emission history generator (MPHG) based on SimSET/PHG to support a majority of the medically important positron emitters. We incorporated the new generator codes inside GATE to improve the simulation efficiency of voxelized phantoms in GATE, while eliminating the need for the temporal photon history file. The validation of this new code based on a MicroPET R4 system was conducted for (124)I and (18)F with mouse-like and rat-like phantoms. Comparison of GATE/MPHG with GATE/GEANT4 indicated there is a slight difference in energy spectra for energy below 50 keV due to the lack of x-ray simulation from (124)I decay in the new code. The spatial resolution, scatter fraction and count rate performance are in good agreement between the two codes. For the case studies of (18)F-NaF ((124)I-IAZG) using MOBY phantom with 1  ×  1 × 1 mm(3) voxel sizes, the results show that GATE/MPHG can achieve acceleration factors of approximately 3.1 × (4.5 ×), 6.5 × (10.7 ×) and 9.5 × (31.0 ×) compared with GATE using the regular navigation method, the compressed voxel method and the parameterized tracking technique, respectively. In conclusion, the implementation of MPHG in GATE allows for improved efficiency of voxelized phantom simulations and is suitable for studying clinical and preclinical imaging.

Detection-ability evaluation of the PEImager for positron emission mammography applications.

This work is a pilot study of using a dual-head scanner in positron emission mammograph (PEM). A positron emission imager (PEImager) developed at our laboratory was used as a PEM prototype to obtain data. Dual-head projection imaging mode was used in the PEM study. An iterative algebraic reconstruction was employed to reconstruct projection data to obtain tomograms. A cylindrizal phantom filled with water was applied to simulate a breast and five hollow spheres (2 mm-10 mm diameters) filled with F-18 fluoride simulated tumors in the breast phantom. Preliminary data revealed that the locations and sizes of the hot spots in the breast phantom were determined from the reconstructed images. The ability to detect the tumor embedded in the radioactive water was evaluated. At a tumor-to-normal tissue ratio 20:1, a 3 mm tumor was detected; 5 mm and 10 mm tumors could be detected at the ratios of 10:1 and 5:1, respectively.

A three-dimensional registration method for automated fusion of micro PET-CT-SPECT whole-body images.

Micro positron emission tomography (PET) and micro single-photon emission computed tomography (SPECT), used for imaging small animals, have become essential tools in developing new pharmaceuticals and can be used, among other things, to test new therapeutic approaches in animal models of human disease, as well as to image gene expression. These imaging techniques can be used noninvasively in both detection and quantification. However, functional images provide little information on the structure of tissues and organs, which makes the localization of lesions difficult. Image fusion techniques can be exploited to map the functional images to structural images, such as X-ray computed tomography (CT), to support target identification and to facilitate the interpretation of PET or SPECT studies. Furthermore, the mapping of two functional images of SPECT and PET on a structural CT image can be beneficial for those in vivo studies that require two biological processes to be monitored simultaneously. This paper proposes an automated method for registering PET, CT, and SPECT images for small animals. A calibration phantom and a holder were used to determine the relationship among three-dimensional fields of view of various modalities. The holder was arranged in fixed positions on the couches of the scanners, and the spatial transformation matrix between the modalities was held unchanged. As long as objects were scanned together with the holder, the predetermined matrix could register the acquired tomograms from different modalities, independently of the imaged objects. In this work, the PET scan was performed by Concorde's microPET R4 scanner, and the SPECT and CT data were obtained using the Gamma Medica's X-SPECT/CT system. Fusion studies on phantoms and animals have been successfully performed using this method. For microPET-CT fusion, the maximum registration errors were 0.21 mm +/- 0.14 mm, 0.26 mm +/- 0.14 mm, and 0.45 mm +/- 0.34 mm in the X (right-left), Y (upper lower), and Z (rostral-caudal) directions, respectively; for the microPET-SPECT fusion, they were 0.24 mm +/- 0.14 mm, 0.28 mm +/- 0.15 mm, and 0.54 mm +/- 0.35 mm in the X, Y, and Z directions, respectively. The results indicate that this simple method can be used in routine fusion studies.