Deep neural network improves fracture detection by clinicians

Lindsey, Robert; Daluiski, Aaron; Chopra, Sumit; Lachapelle, Alexander; Mozer, Michael C.; Sicular, Serge; Hanel, Douglas P.; Gardner, Michael J.; Gupta, Anurag; Hotchkiss, Robert; Potter, Hollis G.

doi:10.1073/pnas.1806905115

Cited by 451 publications

(362 citation statements)

References 25 publications

Supporting

Mentioning

323

Contrasting

Unclassified

Order By: Relevance

“…We did not show shell color by reflection spectra of visible light, except for M. nobilis, so that we could integrate our data with those of previous studies, in which shell color was given simply by color names. Color recognition, however, differs between countries, cultures, and even between individuals [44] . For example, we distinguished three colors resembling each other (maroon, purple, and pink) according to previous studies [9,19] , but no significant differences were observed in the peak positions of ν 1 and ν 2 between those categories (Table 2, Figure 6).…”

Section: Conjugated Length Of Polyenesmentioning

confidence: 99%

Chemical basis of molluscan shell colors revealed with in situ micro‐Raman spectroscopy

Ishikawa

Kagi

Sasaki

et al. 2019

J Raman Spectroscopy

View full text Add to dashboard Cite

Molluscan shells exhibit a wide variety of color tones and patterns that are thought to increase fitness in different natural habitats. However, the chemical aspects of shell pigments have not been well understood. We studied the chemical basis of molluscan shell colors via resonance Raman microspectroscopy using a 514.5‐nm laser on six gastropod, 15 bivalve, and one cephalopod species. Polyene specific signals for C═C (ν1) or C─C (ν2) stretching bands were detected principally from the pigmented regions, and those peak positions correlated with the pigment colors. Because Raman shifts of those bands depend on the conjugation length of polyene backbones, which in turn, affect the highest occupied molecular orbital/lowest unoccupied molecular orbital energy gap, our observations suggest that the conjugation length of polyene is a determinant factor of shell color. Further, our observations and data from previous studies seemed to reveal a quantitative relationship between polyene backbone length and shell color. From an analysis of 89 polyene‐based pigments in 45 molluscan species, we found that yellow, orange/red, and purple/brown/black pigments originate from polyenes with conjugation lengths of 7–9, 9–10, and 10–12, respectively. The shell color shifts from yellow, orange, and red to purple as the polyene chain length increases as reported for free carotenoids. The biological mechanisms underlying the spatial regulations of polyene conjugation systems in molluscan shells will be the most important and intriguing issue to be addressed in the future.

show abstract

Section: Conjugated Length Of Polyenesmentioning

confidence: 99%

Chemical basis of molluscan shell colors revealed with in situ micro‐Raman spectroscopy

Ishikawa

Kagi

Sasaki

et al. 2019

J Raman Spectroscopy

View full text Add to dashboard Cite

show abstract

“…Artificial neural networks exploit a stacked architecture of layers of "neurons" to learn hierarchical representations of data across multiple levels of abstraction, calculating more and more complex features in each layer. Convolutional neural networks, the standard in computer vision, use sets of filters in each layer to generate many complex features from an input image and have shown great promise in many areas of radiography, including in many musculoskeletal applications (10,11,12,13,14,15,16).…”

mentioning

confidence: 99%

Automatic Hip Fracture Identification and Functional Subclassification with Deep Learning

Krogue

Cheng

Hwang³

et al. 2020

Radiology: Artificial Intelligence

113

View full text Add to dashboard Cite

Purpose: Hip fractures are a common cause of morbidity and mortality. Automatic identification and classification of hip fractures using deep learning may improve outcomes by reducing diagnostic errors and decreasing time to operation. Methods: Hip and pelvic radiographs from 1118 studies were reviewed and 3034 hips were labeled via bounding boxes and classified as normal, displaced femoral neck fracture, nondisplaced femoral neck fracture, intertrochanteric fracture, previous ORIF, or previous arthroplasty. A deep learning-based object detection model was trained to automate the placement of the bounding boxes. A Densely Connected Convolutional Neural Network (DenseNet) was trained on a subset of the bounding box images, and its performance evaluated on a held out test set and by comparison on a 100-image subset to two groups of human observers: fellowshiptrained radiologists and orthopaedists, and senior residents in emergency medicine, radiology, and orthopaedics. Results: The binary accuracy for fracture of our model was 93.8% (95% CI, 91.3-95.8%), with sensitivity of 92.7% (95% CI, 88.7-95.6%), and specificity 95.0% (95% CI, 91.5-97.3%). Multiclass classification accuracy was 90.4% (95% CI, 87.4-92.9%). When compared to human observers, our model achieved at least expert-level classification under all conditions. Additionally, when the model was used as an aid, human performance improved, with aided resident performance approximating unaided fellowship-trained expert performance. Conclusions: Our deep learning model identified and classified hip fractures with at least expert-level accuracy, and when used as an aid improved human performance, with aided resident performance approximating that of unaided fellowship-trained attendings.

show abstract

“…Kim and MacKinnon (20) and Lindsey and colleagues (21) used deep convolutional neural networks to assess wrist radiographs for fractures. Kim and MacKinnon attained a sensitivity of 0.9 and specificity of 0.88 in wrist fracture detection with transfer learning on a deep CNN model pretrained on non-medical images.…”

Section: Radiographymentioning

confidence: 99%

Artificial Intelligence in Musculoskeletal Imaging: A Paradigm Shift

Burns

Yao

Summers

2019

Journal of Bone and Mineral Research

View full text Add to dashboard Cite

Artificial intelligence is upending many of our assumptions about the ability of computers to detect and diagnose diseases on medical images. Deep learning, a recent innovation in artificial intelligence, has shown the ability to interpret medical images with sensitivities and specificities at or near that of skilled clinicians for some applications. In this review, we summarize the history of artificial intelligence, present some recent research advances, and speculate about the potential revolutionary clinical impact of the latest computer techniques for bone and muscle imaging. © 2019 American Society for Bone and Mineral Research. Published 2019. This article is a U.S. Government work and is in the public domain in the USA.

show abstract

Deep neural network improves fracture detection by clinicians

Cited by 451 publications

References 25 publications

Chemical basis of molluscan shell colors revealed with in situ micro‐Raman spectroscopy

Chemical basis of molluscan shell colors revealed with in situ micro‐Raman spectroscopy

Automatic Hip Fracture Identification and Functional Subclassification with Deep Learning

Artificial Intelligence in Musculoskeletal Imaging: A Paradigm Shift

Contact Info

Product

Resources

About