TensorFlow: Biology’s Gateway to Deep Learning?

Rampášek, Ladislav; Goldenberg, Anna

doi:10.1016/j.cels.2016.01.009

Cited by 190 publications

(130 citation statements)

References 6 publications

Supporting

Mentioning

128

Contrasting

Unclassified

Order By: Relevance

“…Training was performed with 1000 epochs, needing ∼20 hours of training using a single GPU (NVIDIA TITAN XP). For both training and testing the neural network, we used Keras framework with TensorFlow backend, CUDA8 and CUDNN5.1, on a Linux server.…”

Section: Methodsmentioning

confidence: 99%

Data‐driven synthetic MRI FLAIR artifact correction via deep neural network

Ryu

Nam

Gho³

et al. 2019

Magnetic Resonance Imaging

View full text Add to dashboard Cite

Background FLAIR (fluid attenuated inversion recovery) imaging via synthetic MRI methods leads to artifacts in the brain, which can cause diagnostic limitations. The main sources of the artifacts are attributed to the partial volume effect and flow, which are difficult to correct by analytical modeling. In this study, a deep learning (DL)‐based synthetic FLAIR method was developed, which does not require analytical modeling of the signal. Purpose To correct artifacts in synthetic FLAIR using a DL method. Study Type Retrospective. Subjects A total of 80 subjects with clinical indications (60.6 ± 16.7 years, 38 males, 42 females) were divided into three groups: a training set (56 subjects, 62.1 ± 14.8 years, 25 males, 31 females), a validation set (1 subject, 62 years, male), and the testing set (23 subjects, 57.3 ± 20.4 years, 13 males, 10 females). Field Strength/Sequence 3 T MRI using a multiple‐dynamic multiple‐echo acquisition (MDME) sequence for synthetic MRI and a conventional FLAIR sequence. Assessment Normalized root mean square (NRMSE) and structural similarity (SSIM) were computed for uncorrected synthetic FLAIR and DL‐corrected FLAIR. In addition, three neuroradiologists scored the three FLAIR datasets blindly, evaluating image quality and artifacts for sulci/periventricular and intraventricular/cistern space regions. Statistical Tests Pairwise Student's t‐tests and a Wilcoxon test were performed. Results For quantitative assessment, NRMSE improved from 4.2% to 2.9% (P < 0.0001) and SSIM improved from 0.85 to 0.93 (P < 0.0001). Additionally, NRMSE values significantly improved from 1.58% to 1.26% (P < 0.001), 3.1% to 1.5% (P < 0.0001), and 2.7% to 1.4% (P < 0.0001) in white matter, gray matter, and cerebral spinal fluid (CSF) regions, respectively, when using DL‐corrected FLAIR. For qualitative assessment, DL correction achieved improved overall quality, fewer artifacts in sulci and periventricular regions, and in intraventricular and cistern space regions. Data Conclusion The DL approach provides a promising method to correct artifacts in synthetic FLAIR. Level of Evidence: 4 Technical Efficacy: Stage 1 J. Magn. Reson. Imaging 2019;50:1413–1423.

show abstract

Section: Methodsmentioning

confidence: 99%

Data‐driven synthetic MRI FLAIR artifact correction via deep neural network

Ryu

Nam

Gho³

et al. 2019

Magnetic Resonance Imaging

View full text Add to dashboard Cite

show abstract

“…Deep learning frameworks Deep learning frameworks have been developed to easily build neural networks from existing modules on a high level. The most popular ones are Caffe (Jia et al, 2014), Theano (Bastien et al, 2012), Torch7 (Collobert et al, 2011) and TensorFlow (Abadi et al, 2016;Rampasek & Goldenberg, 2016) (Table 1), which differ in modularity, ease of use and the way models are defined and trained.…”

Section: Off-the-shelf Tools and Practical Considerationsmentioning

confidence: 99%

Deep learning for computational biology

Angermueller

Pärnamaa

Parts

2016

Molecular Systems Biology

1,163

868

View full text Add to dashboard Cite

Technological advances in genomics and imaging have led to an explosion of molecular and cellular profiling data from large numbers of samples. This rapid increase in biological data dimension and acquisition rate is challenging conventional analysis strategies. Modern machine learning methods, such as deep learning, promise to leverage very large data sets for finding hidden structure within them, and for making accurate predictions. In this review, we discuss applications of this new breed of analysis approaches in regulatory genomics and cellular imaging. We provide background of what deep learning is, and the settings in which it can be successfully applied to derive biological insights. In addition to presenting specific applications and providing tips for practical use, we also highlight possible pitfalls and limitations to guide computational biologists when and how to make the most use of this new technology.

show abstract

“…For image recognition, the first layers of the neural network process basic information such as lines and curves while the higher layers in the hierarchy process complex and abstract information. 18 All layers except the final layer of this neural network are pre-trained with more than 1.2 million images. The final layer of the neural network was retrained with the gathered dermatological images.…”

Section: Inception Version 3 (V-3) Inception (V-3) Is a Deep Convolumentioning

confidence: 99%

Data augmentation in dermatology image recognition using machine learning

Aggarwal

2019

Skin Res Technol

View full text Add to dashboard Cite

Background: Each year in the United States, over 80 million people are affected by acne, atopic dermatitis, rosacea, psoriasis, and impetigo. Artificial intelligence and machine learning could prove to be a good tool for assisting in the diagnosis of dermatological conditions. The objective of this study was to evaluate the use of data augmentation in machine learning image recognition of five dermatological disease manifestations-acne, atopic dermatitis, impetigo, psoriasis, and rosacea. Materials and Methods:Open-source dermatological images were gathered and used to retrain TensorFlow Inception version-3. Retraining was done twice-once with and once without data augmentation. Both models were tested with the same images, and R software was used to perform statistical analysis. Results:The average of each of the statistical measures (sensitivity, specificity, PPV, NPN, MCC, and F1 Score) increased when data augmentation was added to the model. In particular, the average Matthews correlation coefficient increased by 7.7%.Each of the five dermatological manifestations had an increase in area under the curve (AUC) after data augmentation with the average increase in AUC of 0.132 and a standard deviation of 0.033. Atopic dermatitis had the highest increase in AUC of 0.18. With data augmentation, the lowest AUC was 0.87 for psoriasis and the highest was 0.97 for acne, indicating that the model performs well. Conclusion:With a deep learning-based approach, it is possible to differentiate dermatological images with appreciable MCC, F1 score, and AUC. Further, data augmentation can be used to increase the model's accuracy by a significant amount. K E Y W O R D Sacne, artificial intelligence, atopic dermatitis, image recognition, impetigo, psoriasis, rosacea How to cite this article: Aggarwal 1LP. Data augmentation in dermatology image recognition using machine learning. Skin

show abstract

TensorFlow: Biology’s Gateway to Deep Learning?

Abstract: TensorFlow is Google's recently released open-source software for deep learning. What are its applications for computational biology?

Cited by 190 publications

References 6 publications

Data‐driven synthetic MRI FLAIR artifact correction via deep neural network

Data‐driven synthetic MRI FLAIR artifact correction via deep neural network

Deep learning for computational biology

Data augmentation in dermatology image recognition using machine learning

Contact Info

Product

Resources

About