Artificial intelligence with deep learning in nuclear medicine and radiology

Decuyper, Milan; Maebe, Jens; Holen, Roel Van; Vandenberghe, Stefaan

doi:10.1186/s40658-021-00426-y

Cited by 34 publications

(18 citation statements)

References 182 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Nowadays, artificial intelligence (AI) deep learning is increasingly being introduced to improve the acquisition performance and the image quality of reconstructed images [24], especially for sparse PET configurations [25]. This can be done using a trained convolutional neural network (CNN) to predict from low count reconstructions the high count reconstructions.…”

Section: Discussionmentioning

confidence: 99%

The potential of a medium-cost long axial FOV PET system for nuclear medicine departments

Vandenberghe

Karakatsanis²,

Akl³

et al. 2022

Eur J Nucl Med Mol Imaging

Self Cite

View full text Add to dashboard Cite

Purpose Total body positron emission tomography (TB-PET) has recently been introduced in nuclear medicine departments. There is a large interest in these systems, but for many centers, the high acquisition cost makes it very difficult to justify their current operational budget. Here, we propose medium-cost long axial FOV scanners as an alternative. Methods Several medium-cost long axial FOV designs are described with their advantages and drawbacks. We describe their potential for higher throughput, more cost-effective scanning, a larger group of indications, and novel research opportunities. The wider spread of TB-PET can also lead to the fast introduction of new tracers (at a low dose), new methodologies, and optimized workflows. Conclusions A medium-cost TB-PET would be positioned between the current standard PET-CT and the full TB-PET systems in investment but recapitulate most advantages of full TB-PET. These systems could be more easily justified financially in a standard academic or large private nuclear medicine department and still have ample research options.

show abstract

Section: Discussionmentioning

confidence: 99%

The potential of a medium-cost long axial FOV PET system for nuclear medicine departments

Vandenberghe

Karakatsanis²,

Akl³

et al. 2022

Eur J Nucl Med Mol Imaging

Self Cite

View full text Add to dashboard Cite

show abstract

“…DL has achieved robust performance in medical elds such as radiology, pathology, and dermatology, which are similar to ophthalmology. [24][25][26] In ophthalmology, DL systems have been shown to accurately detect diabetic retinopathy, age-related macular degeneration, cataract, and glaucoma. [27][28][29][30] Brown [31] created a fully-automated DL system for normal/pre-plus/plus disease using convolutional neural network (CNN) based on 5,511 retinal images, AUC achieved 0.98 for plus diseases, compared with senior ophthalmologists, the DL system also showed strong advantages.…”

Section: Recurrence Rate Complicationsmentioning

confidence: 99%

Prediction of the Reactivation of Retinopathy of Prematurity After Anti-VEGF Treatment Using Machine Learning in Small Numbers

Chen²,

Bai

et al. 2022

Preprint

View full text Add to dashboard Cite

Aim To create and validate a prediction model for retinopathy of prematurity (ROP) reactivation after anti-VEGF therapy with clinical risk factors and retinal images. Methods Infants with TR-ROP undergoing anti-VEGF treatment were recruited from two hospitals, and three models were constructed using machine learning and deep learning algorithms. The areas under the curve (AUC), sensitivity (SEN) and specificity (SPC) were used to show the performances of the prediction models. Results Finally, we included 87 cases, including 21 with recurrent and 66 nonrecurrent cases. The AUC for the clinical risk factor model was 0.80 and 0.77 in the internal and external validation groups, respectively. The average AUC, sensitivity, and specificity in the internal validation for the retinal image model were 0.82, 0.93, and 0.63, respectively. The SPC, AUC, and SEN for the combined model were 0.73, 0.84, and 0.93, separately. Conclusion We constructed a prediction model for the reactivation of ROP. Using this prediction model, we can optimize strategies for treating TR-TOP infants and developing screening plans after treatment.

show abstract

“…Smoothing lters used to reduce image noise have the tradeoff of degrading resolution. Additionally, many studies have reported improving nuclear medicine image quality using deep learning [9][10][11][12][13][14][15]. These reports mainly used tomographic imaging techniques, such as SPECT and positron emission tomography.…”

Section: Introductionmentioning

confidence: 99%

Improvement of Image Quality in Low-Count Bone Scintigraphy Using Deep Learning

Murata

Onoguchi

Hashimoto

et al. 2023

Preprint

View full text Add to dashboard Cite

Objective To improve image quality for low-count bone scintigraphy whole-body images using deep learning and evaluate their applicability in clinical practice.Methods Five hundred fifty patients were included in the study. Low-count Original images (75%, 50%, 25%, 10%, and 5% counts) were generated from Reference images (100% counts) using Poisson resampling. Patients were randomly divided into training (500) and evaluation (50) groups. Output (DL-filtered) images were obtained after training with U-Net using Reference images as teacher data. Gaussian-filtered images were generated for comparison. Peak signal-to-noise ratio (PSNR) and structural similarity (SSIM) to the Reference image were calculated to determine image quality. Artificial neural network (ANN) value, bone scan index (BSI), and number of hotspots (Hs) were computed using BONENAVI analysis for patients with and without bone metastases, to assess diagnostic performance. Accuracy of bone metastasis detection and area under the curve (AUC) were calculated. Original, Gaussian-filtered, and DL-filtered images were compared with Reference images.Results PSNR and SSIM for DL-filtered images were highest in all count percentages. BONENAVI analysis values for DL-filtered images did not differ significantly regardless of the presence or absence of bone metastases. BONENAVI analysis values for Original and Gaussian-filtered images differed significantly at < 25% counts in patients without bone metastases. In patients with bone metastases, BSI and Hs for Original and Gaussian-filtered images differed significantly at < 10% counts, whereas ANN values did not. Accuracy of bone metastasis detection was highest for DL-filtered images in all count percentages; AUC did not differ significantly. Accuracy of Original and Gaussian-filtered images worsened with decreasing count percentage; AUC differed significantly for Original images at < 50% counts and for Gaussian-filtered images at < 25% counts.Conclusions Our deep learning model improved image quality and bone metastasis detection accuracy for low-count bone scintigraphy whole-body images, suggesting its applicability in clinical practice.

show abstract

Artificial intelligence with deep learning in nuclear medicine and radiology

Cited by 34 publications

References 182 publications

The potential of a medium-cost long axial FOV PET system for nuclear medicine departments

The potential of a medium-cost long axial FOV PET system for nuclear medicine departments

Prediction of the Reactivation of Retinopathy of Prematurity After Anti-VEGF Treatment Using Machine Learning in Small Numbers

Improvement of Image Quality in Low-Count Bone Scintigraphy Using Deep Learning

Contact Info

Product

Resources

About