Leveraging deep neural networks to improve numerical and perceptual image quality in low-dose preclinical PET imaging

A guide to artificial intelligence for cancer researchers

Artificial intelligence (AI) has been commoditized. It has evolved from a specialty resource to a readily accessible tool for cancer researchers. AI-based tools can boost research productivity in daily workflows, but can also extract hidden information from existing data, thereby enabling new scientific discoveries. Building a basic literacy in these tools is useful for every cancer researcher. Researchers with a traditional biological science focus can use AI-based tools through off-the-shelf software, whereas those who are more computationally inclined can develop their own AI-based software pipelines. In this article, we provide a practical guide for non-computational cancer researchers to understand how AI-based tools can benefit them. We convey general principles of AI for applications in image analysis, natural language processing and drug discovery. In addition, we give examples of how non-computational researchers can get started on the journey to productively use AI in their own work.

Machine learning diagnosis of active Juvenile Idiopathic Arthritis on blood pool [ 99M Tc] Tc-MDP scintigraphy images

PURPOSE

Neural network has widely been applied for medical classifications and disease diagnosis. This study employs deep learning to best discriminate Juvenile Idiopathic Arthritis (JIA), a pediatric chronic joint inflammatory disease, from healthy joints by exploring blood pool images of 2phase [ 99m Tc] Tc-MDP bone scintigraphy.

METHODS

Self-deigned multi-input Convolutional Neural Network (CNN) in addition to three available pre-trained models including VGG16, ResNet50 and Xception are applied on 1304 blood pool images of 326 healthy and known JIA children and adolescents (aged 1-16).

RESULTS

The self-designed model ROC analysis shows diagnostic efficiency with Area Under the Curve (AUC) 0.82 and 0.86 for knee and ankle joints, respectively. Among the three pertained models, VGG16 ROC analysis reveals AUC 0.76 and 0.81 for knee and ankle images, respectively.

CONCLUSION

The self-designed model shows best performance on blood pool scintigraph diagnosis of patients with JIA. VGG16 was the most efficient model rather to other pre-trained networks. This study can pave the way of artificial intelligence (AI) application in nuclear medicine for the diagnosis of pediatric inflammatory disease.

Image Denoising of Low-Dose PET Mouse Scans with Deep Learning: Validation Study for Preclinical Imaging Applicability

PURPOSE

Positron emission tomography (PET) image quality can be improved by higher injected activity and/or longer acquisition time, but both may often not be practical in preclinical imaging. Common preclinical radioactive doses (10 MBq) have been shown to cause deterministic changes in biological pathways. Reducing the injected tracer activity and/or shortening the scan time inevitably results in low-count acquisitions which poses a challenge because of the inherent noise introduction. We present an image-based deep learning (DL) framework for denoising lower count micro-PET images.

PROCEDURES

For 36 mice, a 15-min [18F]FDG (8.15 ± 1.34 MBq) PET scan was acquired at 40 min post-injection on the Molecubes β-CUBE (in list mode). The 15-min acquisition (high-count) was parsed into smaller time fractions of 7.50, 3.75, 1.50, and 0.75 min to emulate images reconstructed at 50, 25, 10, and 5% of the full counts, respectively. A 2D U-Net was trained with mean-squared-error loss on 28 high-low count image pairs.

RESULTS

The DL algorithms were visually and quantitatively compared to spatial and edge-preserving denoising filters; the DL-based methods effectively removed image noise and recovered image details much better while keeping quantitative (SUV) accuracy. The largest improvement in image quality was seen in the images reconstructed with 10 and 5% of the counts (equivalent to sub-1 MBq or sub-1 min mouse imaging). The DL-based denoising framework was also successfully applied on the NEMA-NU4 phantom and different tracer studies ([18F]PSMA, [18F]FAPI, and [68 Ga]FAPI).

CONCLUSION

Visual and quantitative results support the superior performance and robustness in image denoising of the implemented DL models for low statistics micro-PET. This offers much more flexibility in optimizing preclinical, longitudinal imaging protocols with reduced tracer doses or shorter durations.

Utilizing deep learning techniques to improve image quality and noise reduction in preclinical low-dose PET images in the sinogram domain

BACKGROUND

Low-dose positron emission tomography (LD-PET) imaging is commonly employed in preclinical research to minimize radiation exposure to animal subjects. However, LD-PET images often exhibit poor quality and high noise levels due to the low signal-to-noise ratio. Deep learning (DL) techniques such as generative adversarial networks (GANs) and convolutional neural network (CNN) have the capability to enhance the quality of images derived from noisy or low-quality PET data, which encodes critical information about radioactivity distribution in the body.

PURPOSE

Our objective was to optimize the image quality and reduce noise in preclinical PET images by utilizing the sinogram domain as input for DL models, resulting in improved image quality as compared to LD-PET images.

METHODS

A GAN and CNN model were utilized to predict high-dose (HD) preclinical PET sinograms from the corresponding LD preclinical PET sinograms. In order to generate the datasets, experiments were conducted on micro-phantoms, animal subjects (rats), and virtual simulations. The quality of DL-generated images was weighted by performing the following quantitative measures: structural similarity index measure (SSIM), root mean squared error (RMSE), peak signal-to-noise ratio (PSNR), signal-to-noise ratio (SNR), and contrast-to-noise ratio (CNR). Additionally, DL input and output were both subjected to a spatial resolution calculation of full width half maximum (FWHM) and full width tenth maximum (FWTM). DL outcomes were then compared with the conventional denoising algorithms such as non-local means (NLM), block-matching, and 3D filtering (BM3D).

RESULTS

The DL models effectively learned image features and produced high-quality images, as reflected in the quantitative metrics. Notably, the FWHM and FWTM values of DL PET images exhibited significantly improved accuracy compared to LD, NLM, and BM3D PET images, and just as precise as HD PET images. The MSE loss underscored the excellent performance of the models, indicating that the models performed well. To further improve the training, the generator loss (G loss) was increased to a value higher than the discriminator loss (D loss), thereby achieving convergence in the GAN model.

CONCLUSIONS

The sinograms generated by the GAN network closely resembled real HD preclinical PET sinograms and were more realistic than LD. There was a noticeable improvement in image quality and noise factor in the predicted HD images. Importantly, DL networks did not fully compromise the spatial resolution of the images.

A neural network based framework for effective laparoscopic video quality assessment

Video quality assessment is a challenging problem having a critical significance in the context of medical imaging. For instance, in laparoscopic surgery, the acquired video data suffers from different kinds of distortion that not only hinder surgery performance but also affect the execution of subsequent tasks in surgical navigation and robotic surgeries. For this reason, we propose in this paper neural network-based approaches for distortion classification as well as quality prediction. More precisely, a Residual Network (ResNet) based approach is firstly developed for simultaneous ranking and classification task. Then, this architecture is extended to make it appropriate for the quality prediction task by using an additional Fully Connected Neural Network (FCNN). To train the overall architecture (ResNet and FCNN models), transfer learning and end-to-end learning approaches are investigated. Experimental results, carried out on a new laparoscopic video quality database, have shown the efficiency of the proposed methods compared to recent conventional and deep learning based approaches.