Effects of Image Size on Deep Learning

Rukundo, Olivier

doi:10.3390/electronics12040985

Cited by 42 publications

(25 citation statements)

References 48 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Our investigation is consistent with previous literature that suggests higher image resolutions yield more accurate results [ 21 , [40] , [41] , [42] ]. However, most literature has focused on training deep learning networks or reducing image resolution to unrecognizable levels to assess accuracy and loss coefficients.…”

Section: Discussionsupporting

confidence: 93%

“…However, traditional image processing methods have not been discussed or compared, and only a few studies have examined the effect of image datasets on the analysis results during deep learning network training. Many studies have used thousands of images from the internet [ 39 ] or generated additional images through image rotation and scaling for network training [ 30 , 42 ], but they have ignored the correlation between the training and analysis images and the challenges associated with obtaining images from experimental fields.…”

Section: Discussionmentioning

confidence: 99%

See 1 more Smart Citation

Evaluation of deep learning computer vision for water level measurements in rivers

Liu,

Huang

2024

Heliyon

View full text Add to dashboard Cite

Section: Discussionsupporting

confidence: 93%

Section: Discussionmentioning

confidence: 99%

Evaluation of deep learning computer vision for water level measurements in rivers

Liu,

Huang

2024

Heliyon

View full text Add to dashboard Cite

“…Second, the size of each image was small. These limitations might affect the image quality, although image size reduction is inevitable because of lower computational resources [ 45 ]. Therefore, the image quality of PET GE-FDG is insufficient for visual assessments in daily practice.…”

Section: Discussionmentioning

confidence: 99%

Generation of Conventional 18F-FDG PET Images from 18F-Florbetaben PET Images Using Generative Adversarial Network: A Preliminary Study Using ADNI Dataset

et al. 2023

View full text Add to dashboard Cite

Background and Objectives: 18F-fluorodeoxyglucose (FDG) positron emission tomography (PET) (PETFDG) image can visualize neuronal injury of the brain in Alzheimer’s disease. Early-phase amyloid PET image is reported to be similar to PETFDG image. This study aimed to generate PETFDG images from 18F-florbetaben PET (PETFBB) images using a generative adversarial network (GAN) and compare the generated PETFDG (PETGE-FDG) with real PETFDG (PETRE-FDG) images using the structural similarity index measure (SSIM) and the peak signal-to-noise ratio (PSNR). Materials and Methods: Using the Alzheimer’s Disease Neuroimaging Initiative (ADNI) database, 110 participants with both PETFDG and PETFBB images at baseline were included. The paired PETFDG and PETFBB images included six and four subset images, respectively. Each subset image had a 5 min acquisition time. These subsets were randomly sampled and divided into 249 paired PETFDG and PETFBB subset images for the training datasets and 95 paired subset images for the validation datasets during the deep-learning process. The deep learning model used in this study is composed of a GAN with a U-Net. The differences in the SSIM and PSNR values between the PETGE-FDG and PETRE-FDG images in the cycleGAN and pix2pix models were evaluated using the independent Student’s t-test. Statistical significance was set at p ≤ 0.05. Results: The participant demographics (age, sex, or diagnosis) showed no statistically significant differences between the training (82 participants) and validation (28 participants) groups. The mean SSIM between the PETGE-FDG and PETRE-FDG images was 0.768 ± 0.135 for the cycleGAN model and 0.745 ± 0.143 for the pix2pix model. The mean PSNR was 32.4 ± 9.5 and 30.7 ± 8.0. The PETGE-FDG images of the cycleGAN model showed statistically higher mean SSIM than those of the pix2pix model (p < 0.001). The mean PSNR was also higher in the PETGE-FDG images of the cycleGAN model than those of pix2pix model (p < 0.001). Conclusions: We generated PETFDG images from PETFBB images using deep learning. The cycleGAN model generated PETGE-FDG images with a higher SSIM and PSNR values than the pix2pix model. Image-to-image translation using deep learning may be useful for generating PETFDG images. These may provide additional information for the management of Alzheimer’s disease without extra image acquisition and the consequent increase in radiation exposure, inconvenience, or expenses.

show abstract

“…This is a crucial step because inaccurate or unnecessary data can harm the model's performance [39]. Once the dataset had been cleaned, the next step was to resize the images to a consistent size [40]. In this case, the images were resized to 224x224 pixels [41].…”

Section: Data Preprocessingmentioning

confidence: 99%

Deep learning model enhanced skin cancer detection

Kolachina

Agada²,

Li³

et al. 2023

Biophotonics and Immune Responses XVIII

View full text Add to dashboard Cite

According to the American Academy of Dermatology Association, identifying the types of skin cancer depends on the origin of a cell mutation resulting in the rapid growth of these abnormal cells in the epidermis. These mutations lead the skin cells to multiply rapidly and form malignant tumors. Skin cancer is ranked as the 17th most common form of cancer worldwide according to the World Cancer Research Fund. Skin cancer treatments cost the United States more than $8 billion (about $25 per person in the US) each year, making skin cancer the fifth most costly cancer for Medicare. Furthermore, skin cancer is an under recognized problem for diverse populations, including young women and minorities. Researchers have been exploring different technologies to detect skin care at its early stage to avoid high mortality rate and expensive medical treatment. This paper presents a novel ensembled deep learning model for the early detection of skin cancer. Our research is based on The HAM10000 dataset, a diverse collection of multi-sourced dermatoscopic images of common pigmented skin lesions which consists of 10015 dermatoscopic images. We have compared several deep learning neural network architectures and classifiers such as DNN, RNN, SVM and KNN in terms of accuracy rate and computation complexity and presented an ensembled deep learning model for early skin cancer detection. The main contribution of this paper is the productions of a comparative study of several skin cancer detection techniques using powerful computer vision techniques and deep learning models and a novel ensembled deep learning model for skin cancer detection.

show abstract

Effects of Image Size on Deep Learning

Cited by 42 publications

References 48 publications

Evaluation of deep learning computer vision for water level measurements in rivers

Evaluation of deep learning computer vision for water level measurements in rivers

Generation of Conventional 18F-FDG PET Images from 18F-Florbetaben PET Images Using Generative Adversarial Network: A Preliminary Study Using ADNI Dataset

Deep learning model enhanced skin cancer detection

Contact Info

Product

Resources

About