Enhancing SNR in CEST imaging: A deep learning approach with a denoising convolutional autoencoder.

PURPOSE: To develop a SNR enhancement method for CEST imaging using a denoising convolutional autoencoder (DCAE) and compare its performance with state-of-the-art denoising methods. METHOD: The DCAE-CEST model encompasses an encoder and a decoder network. The encoder learns features from the input CEST Z-spectrum via a series of one-dimensional convolutions, nonlinearity applications, and pooling. Subsequently, the decoder reconstructs an output denoised Z-spectrum using a series of up-sampling and convolution layers. The DCAE-CEST model underwent multistage training in an environment constrained by Kullback-Leibler divergence, while ensuring data adaptability through context learning using Principal Component Analysis-processed Z-spectrum as a reference. The model was trained using simulated Z-spectra, and its performance was evaluated using both simulated data and in vivo data from an animal tumor model. Maps of amide proton transfer (APT) and nuclear Overhauser enhancement (NOE) effects were quantified using the multiple-pool Lorentzian fit, along with an apparent exchange-dependent relaxation metric. RESULTS: In digital phantom experiments, the DCAE-CEST method exhibited superior performance, surpassing existing denoising techniques, as indicated by the peak SNR and Structural Similarity Index. Additionally, in vivo data further confirm the effectiveness of the DCAE-CEST in denoising the APT and NOE maps when compared with other methods. Although no significant difference was observed in APT between tumors and normal tissues, there was a significant difference in NOE, consistent with previous findings. CONCLUSION: The DCAE-CEST can learn the most important features of the CEST Z-spectrum and provide the most effective denoising solution compared with other methods.

A Denoising Convolutional Autoencoder for SNR Enhancement in Chemical Exchange Saturation Transfer imaging: (DCAE-CEST).

Purpose: To develop a SNR enhancement method for chemical exchange saturation transfer (CEST) imaging using a denoising convolutional autoencoder (DCAE), and compare its performance with state-of-the-art denoising methods. Method: The DCAE-CEST model encompasses an encoder and a decoder network. The encoder learns features from the input CEST Z-spectrum via a series of 1D convolutions, nonlinearity applications and pooling. Subsequently, the decoder reconstructs an output denoised Z-spectrum using a series of up-sampling and convolution layers. The DCAE-CEST model underwent multistage training in an environment constrained by Kullback-Leibler divergence, while ensuring data adaptability through context learning using Principal Component Analysis processed Z-spectrum as a reference. The model was trained using simulated Z-spectra, and its performance was evaluated using both simulated data and in-vivo data from an animal tumor model. Maps of amide proton transfer (APT) and nuclear Overhauser enhancement (NOE) effects were quantified using the multiple-pool Lorentzian fit, along with an apparent exchange-dependent relaxation metric. Results: In digital phantom experiments, the DCAE-CEST method exhibited superior performance, surpassing existing denoising techniques, as indicated by the peak SNR and Structural Similarity Index. Additionally, in vivo data further confirms the effectiveness of the DCAE-CEST in denoising the APT and NOE maps when compared to other methods. While no significant difference was observed in APT between tumors and normal tissues, there was a significant difference in NOE, consistent with previous findings. Conclusion: The DCAE-CEST can learn the most important features of the CEST Z-spectrum and provide the most effective denoising solution compared to other methods.

A rapid method for phosphocreatine-weighted imaging in muscle using double saturation power-chemical exchange saturation transfer.

Monitoring the variation in phosphocreatine (PCr) levels following exercise provides valuable insights into muscle function. Chemical exchange saturation transfer (CEST) has emerged as a sensitive method with which to measure PCr levels in muscle, surpassing conventional MR spectroscopy. However, existing approaches for quantifying PCr CEST signals rely on time-consuming fitting methods that require the acquisition of the entire or a section of the CEST Z-spectrum. Additionally, traditional fitting methods often necessitate clear CEST peaks, which may be challenging to obtain at low magnetic fields. This paper evaluated the application of a new model-free method using double saturation power (DSP), termed DSP-CEST, to estimate the PCr CEST signal in muscle. The DSP-CEST method requires the acquisition of only two or a few CEST signals at the PCr frequency offset with two different saturation powers, enabling rapid dynamic imaging. Additionally, the DSP-CEST approach inherently eliminates confounding signals, offering enhanced robustness compared with fitting methods. Furthermore, DSP-CEST does not demand clear CEST peaks, making it suitable for low-field applications. We evaluated the capability of DSP-CEST to enhance the specificity of PCr CEST imaging through simulations and experiments on muscle tissue phantoms at 4.7 T. Furthermore, we applied DSP-CEST to animal leg muscle both before and after euthanasia and observed successful reduction of confounding signals. The DSP-CEST signal still has contaminations from a residual magnetization transfer (MT) effect and an aromatic nuclear Overhauser enhancement effect, and thus only provides a PCr-weighted imaging. The residual MT effect can be reduced by a subtraction of DSP-CEST signals at 2.6 and 5 ppm. Results show that the residual MT-corrected DSP-CEST signal at 2.6 ppm has significant variation in postmortem tissues. By contrast, both the CEST signal at 2.6 ppm and a conventional Lorentzian difference analysis of CEST signal at 2.6 ppm demonstrate no significant variation in postmortem tissues.

Machine learning-based amide proton transfer imaging using partially synthetic training data.

PURPOSE: Machine learning (ML) has been increasingly used to quantify CEST effect. ML models are typically trained using either measured data or fully simulated data. However, training with measured data often lacks sufficient training data, whereas training with fully simulated data may introduce bias because of limited simulations pools. This study introduces a new platform that combines simulated and measured components to generate partially synthetic CEST data, and to evaluate its feasibility for training ML models to predict amide proton transfer (APT) effect. METHODS: Partially synthetic CEST signals were created using an inverse summation of APT effects from simulations and the other components from measurements. Training data were generated by varying APT simulation parameters and applying scaling factors to adjust the measured components, achieving a balance between simulation flexibility and fidelity. First, tissue-mimicking CEST signals along with ground truth information were created using multiple-pool model simulations to validate this method. Second, an ML model was trained individually on partially synthetic data, in vivo data, and fully simulated data, to predict APT effect in rat brains bearing 9 L tumors. RESULTS: Experiments on tissue-mimicking data suggest that the ML method using the partially synthetic data is accurate in predicting APT. In vivo experiments suggest that our method provides more accurate and robust prediction than the training using in vivo data and fully synthetic data. CONCLUSION: Partially synthetic CEST data can address the challenges in conventional ML methods.

Nuclear Overhauser enhancement imaging at -1.6 ppm in rat brain at 4.7T.

PURPOSE: A new nuclear Overhauser enhancement (NOE)-mediated saturation transfer signal at around -1.6 ppm, termed NOE(-1.6), has been reported at high fields of 7T and 9.4T previously. This study aims to validate the presence of this signal at a relatively low field of 4.7T and evaluate its variations in different brain regions and tumors. METHODS: Rats were injected with monocrystalline iron oxide nanoparticles to reduce the NOE(-1.6) signal. CEST signals were measured using different saturation powers before and after injection to assess the presence of this signal. Multiple-pool Lorentzian fits, with/without inclusion of the NOE(-1.6) pool, were performed on CEST Z-spectra obtained from healthy rat brains and rats with 9L tumors. These fits aimed to further validate the presence of the NOE(-1.6) signal and quantify its amplitude. RESULTS: The NOE(-1.6) signal exhibited a dramatic change following the injection of monocrystalline iron oxide nanoparticles, confirming its presence at 4.7T. The NOE(-1.6) signal reached its peak at a saturation power of ∼0.75 µT, indicating an optimized power level. The multiple-pool Lorentzian fit without the NOE(-1.6) pool showed higher residuals around -1.6 ppm compared to the fit with this pool, further supporting the presence of this signal. The NOE(-1.6) signal did not exhibit significant variation in the corpus callosum and caudate putamen regions, but it showed a significant decrease in tumors, which aligns with previous findings at 9.4T. CONCLUSION: This study successfully demonstrated the presence of the NOE(-1.6) signal at 4.7T, which provides valuable insights into its potential applications at lower field strengths.

Amide Proton Transfer (APT) imaging in tumor with a machine learning approach using partially synthetic data.

Purpose: Machine learning (ML) has been increasingly used to quantify chemical exchange saturation transfer (CEST) effect. ML models are typically trained using either measured data or fully simulated data. However, training with measured data often lacks sufficient training data, while training with fully simulated data may introduce bias due to limited simulations pools. This study introduces a new platform that combines simulated and measured components to generate partially synthetic CEST data, and to evaluate its feasibility for training ML models to predict amide proton transfer (APT) effect. Methods: Partially synthetic CEST signals were created using an inverse summation of APT effects from simulations and the other components from measurements. Training data were generated by varying APT simulation parameters and applying scaling factors to adjust the measured components, achieving a balance between simulation flexibility and fidelity. First, tissue-mimicking CEST signals along with ground truth information were created using multiple-pool model simulations to validate this method. Second, an ML model was trained individually on partially synthetic data, in vivo data, and fully simulated data, to predict APT effect in rat brains bearing 9L tumors. Results: Experiments on tissue-mimicking data suggest that the ML method using the partially synthetic data is accurate in predicting APT. In vivo experiments suggest that our method provides more accurate and robust prediction than the training using in vivo data and fully synthetic data. Conclusion: Partially synthetic CEST data can address the challenges in conventional ML methods.

Tumor Malignancy Detection Using Histopathology Imaging.

Image segmentation and classification in the biomedical imaging field has high worth in cancer diagnosis and grading. The proposed method classifies the images based on a combination of handcrafted features and shape features using bag of visual words (BoW). The multistage segmentation technique to localize the nuclei in histopathology images includes the stain decomposition and histogram equalization to highlight the nucleus region, followed by the nuclei key point extraction using fast radial symmetry transform, normalized graph cut based on the nuclei region estimation, and nuclei boundary estimation using modified gradient. Subsequently, features from localized regions termed as handcrafted features and the shape features using BoW are extracted for classification. The experiments are performed using both the handcrafted features and BoW to take the advantages of both local nuclei features and globally spatial features. The simulation is performed on the Bisque and BreakHis data sets (with corresponding average accuracies of 93.87% and 96.96%, respectively) and confirms better diagnosis performance using the proposed method.