Deep Learning in DXA Image Segmentation

Hussain, Dildar; Loh, Woong-Kee; Lee, Jooyoung

doi:10.32604/cmc.2021.013031

Cited by 9 publications

(5 citation statements)

References 42 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…To address this, we implemented transfer learning to enhance the training efficiency of DL models using small femur DXA images. Leveraging the weights of the pre-trained VGG-16 network from the extensive ImageNet dataset, our study on femur segmentation demonstrates that current DL models outperform previously utilized methods [23][24][25][26].…”

Section: Bmd Analysismentioning

confidence: 94%

“…Finally, the segmentation methods' retrieved ground truths (RGT) and manual ground truths (GT) were applied to the test data in each cross-validation fold to compare their performance. For more details, visit our previous work referenced in [23,24,26], as the same evaluation and performance analysis strategy was adopted to evaluate the accuracy of this work.…”

Section: Evaluation and Performance Analysismentioning

confidence: 99%

“…With the inherent limitations described in the references [23][24][25], the existing techniques are unsuitable for DXA image segmentation. All the limitations described in our previous work [26] forced us to new investigations, which led us to more precise segmentation with greater accuracy in DXA imaging.…”

Section: Introductionmentioning

confidence: 97%

See 2 more Smart Citations

Exploring the Impact of Noise and Image Quality on Deep Learning Performance in DXA Images

Hussain,

Hyeon Gu

2024

Diagnostics

View full text Add to dashboard Cite

Background and Objective: Segmentation of the femur in Dual-Energy X-ray (DXA) images poses challenges due to reduced contrast, noise, bone shape variations, and inconsistent X-ray beam penetration. In this study, we investigate the relationship between noise and certain deep learning (DL) techniques for semantic segmentation of the femur to enhance segmentation and bone mineral density (BMD) accuracy by incorporating noise reduction methods into DL models. Methods: Convolutional neural network (CNN)-based models were employed to segment femurs in DXA images and evaluate the effects of noise reduction filters on segmentation accuracy and their effect on BMD calculation. Various noise reduction techniques were integrated into DL-based models to enhance image quality before training. We assessed the performance of the fully convolutional neural network (FCNN) in comparison to noise reduction algorithms and manual segmentation methods. Results: Our study demonstrated that the FCNN outperformed noise reduction algorithms in enhancing segmentation accuracy and enabling precise calculation of BMD. The FCNN-based segmentation approach achieved a segmentation accuracy of 98.84% and a correlation coefficient of 0.9928 for BMD measurements, indicating its effectiveness in the clinical diagnosis of osteoporosis. Conclusions: In conclusion, integrating noise reduction techniques into DL-based models significantly improves femur segmentation accuracy in DXA images. The FCNN model, in particular, shows promising results in enhancing BMD calculation and clinical diagnosis of osteoporosis. These findings highlight the potential of DL techniques in addressing segmentation challenges and improving diagnostic accuracy in medical imaging.

show abstract

Section: Bmd Analysismentioning

confidence: 94%

Section: Evaluation and Performance Analysismentioning

confidence: 99%

See 1 more Smart Citation

Exploring the Impact of Noise and Image Quality on Deep Learning Performance in DXA Images

Hussain,

Hyeon Gu

2024

Diagnostics

View full text Add to dashboard Cite

show abstract

“…In deep learning networks, padding extra pixels to an input is required before convolution in order to obtain an output with the same size of an input, for example in modern networks for segmentation tasks [21][22][23][24][25][26] ZP is used at each convolution layer. Usually padding schemes are paid less attention and simple padding algorithms such as ZP and reflection padding are utilized for padding.…”

Section: Padding Schemesmentioning

confidence: 99%

SRResNet Performance Enhancement Using Patch Inputs and Partial Convolution-Based Padding

Ullah¹,

Song²

2023

Computers, Materials &Amp; Continua

View full text Add to dashboard Cite

Due to highly underdetermined nature of Single Image Super-Resolution (SISR) problem, deep learning neural networks are required to be more deeper to solve the problem effectively. One of deep neural networks successful in the Super-Resolution (SR) problem is ResNet which can render the capability of deeper networks with the help of skip connections. However, zero padding (ZP) scheme in the network restricts benefits of skip connections in SRResNet and its performance as the ratio of the number of pure input data to that of zero padded data increases. In this paper. we consider the ResNet with Partial Convolution based Padding (PCP) instead of ZP to solve SR problem. Since training of deep neural networks using patch images is advantageous in many aspects such as the number of training image data and network complexities, patch image based SR performance is compared with single full image based one. The experimental results show that patch based SRResNet SR results are better than single full image based ones and the performance of deep SRResNet with PCP is better than the one with ZP.

show abstract

“…Osteoporosis detection typically involves a combination of medical history evaluation, physical examination, and imaging tests such as magnetic resonance imaging (MRI), computerized tomography (CT), and dual x-ray absorptiometry (DEXA or DXA) [9] [10]. Through this, the bone mineral density is measured and compared with predefined values (threshold), and according to the result, the case is diagnosed as having osteoporosis or normal.…”

Section: Diagnosis Of Osteoporosis Using Transfer Learning In the Sam...mentioning

confidence: 99%

Diagnosis of Osteoporosis Using Transfer Learning in the Same Domain

Z. Mohammed,

Loay E. George

2023

Int. J. Onl. Eng.

View full text Add to dashboard Cite

This paper presents a system for diagnosing osteoporosis using x-rays by leveraging transfer learning in the same domain. The proposed system consists of phase 1 and phase 2; each phase includes several stages, as the pre-processing stage appropriately prepares the source image via noise reduction by the average filter, contrast enhancement using histogram equalization, and obtaining the region of interest by employing K-mean and edge detection, followed by the smudging stage through a mean filter with a large window size, which subsequently contributed to facilitating the diagnosis. The stages mentioned in both phases are similar. In phase 1, the model is trained on a large unlabeled x-ray dataset collected from different orthopedic centers to identify the general features of the image. In phase 2, fine-tune the trained model with the target dataset; this approach is beneficial when the target task has limited labeled data or when training a model from scratch is computationally expensive. It is worth noting that two datasets were used as target datasets. The accuracy of diagnosing osteoporosis using the proposed deep convolutional neural network (DCNN) model was 94.5 with the osteoporosis knee x-ray database (Dataset A). The accuracy of diagnosing osteoporosis using transfer learning in the same field was 98.91 when training the proposed DCNN model with a large unlabeled dataset and fine-tuning with the target database, osteoporosis knee x-ray database (Dataset A). The accuracy of diagnosing osteoporosis using the proposed DCNN model was 91.5 with the knee x-ray osteoporosis database (Dataset B). The accuracy of diagnosing osteoporosis using transfer learning in the same field was 96.61 when training the proposed DCNN model with a large unlabeled dataset and fine-tuning with the target knee x-ray osteoporosis database (Dataset B).

show abstract

Deep Learning in DXA Image Segmentation

Cited by 9 publications

References 42 publications

Exploring the Impact of Noise and Image Quality on Deep Learning Performance in DXA Images

Exploring the Impact of Noise and Image Quality on Deep Learning Performance in DXA Images

SRResNet Performance Enhancement Using Patch Inputs and Partial Convolution-Based Padding

Diagnosis of Osteoporosis Using Transfer Learning in the Same Domain

Contact Info

Product

Resources

About