Impact of scan quality on AI assessment of hip dysplasia ultrasound

Hareendranathan, Abhilash Rakkunedeth; Chahal, Baljot S.; Ghasseminia, Siyavash; Zonoobi, Dornoosh; Jaremko, Jacob L.

doi:10.1007/s40477-021-00560-4

Cited by 12 publications

(6 citation statements)

References 28 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…AI performance would likely be improved by an initial automated prescreen of images to reject the lowest-quality images from being analyzed. 23 Overall, we observed only slightly lower interobserver reliability on sweeps than on conventional single 2D ultrasound images for hip dysplasia index measurement and diagnosis and found that an AI package interpreted sweeps similarly to expert nonsubspecialist human readers, performing only slightly lower compared with the widely accepted standard. Since sweeps provide more data than single images, are more easily obtained by nonexperts, and are amenable to postscan analysis by human experts or AI, these results motivate further study of hip dysplasia ultrasound using sweeps.…”

Section: Discussionmentioning

confidence: 60%

Interobserver Variability of Hip Dysplasia Indices on Sweep Ultrasound for Novices, Experts, and Artificial Intelligence

Ghasseminia

Lim

Concepcion

et al. 2022

Journal of Pediatric Orthopaedics

Self Cite

View full text Add to dashboard Cite

Background: Ultrasound for developmental dysplasia of the hip (DDH) is challenging for nonexperts to perform and interpret. Recording "sweep" images allows more complete hip assessment, suitable for automation by artificial intelligence (AI), but reliability has not been established. We assessed agreement between readers of varying experience and a commercial AI algorithm, in DDH detection from infant hip ultrasound sweeps. Methods: We selected a full spectrum of poor-to-excellent quality images and normal to severe dysplasia, in 240 hips (120 single 2-dimensional images, 120 sweeps). For 12 readers (radiologists, sonographers, clinicians and researchers; 3 were DDH subspecialists), and a ultrasound-FDA-cleared AI software package (Medo Hip), we calculated interobserver reliability for alpha angle measurements by intraclass correlation coefficient (ICC 2,1 ) and for DDH classification by Randolph Kappa. Results: Alpha angle reliability was high for AI versus subspecialists (ICC = 0.87 for sweeps, 0.90 for single images). For DDH diagnosis from sweeps, agreement was high between subspecialists (kappa = 0.72), and moderate for nonsubspecialists (0.54) and AI (0.47). Agreement was higher for single images (kappa = 0.80, 0.66, 0.49). AI reliability deteriorated more than human readers for the poorest-quality images. The agreement of radiologists and clinicians with the accepted standard, while still high, was significantly poorer for sweeps than 2D images (P < 0.05). Conclusions: In a challenging exercise representing the wide spectrum of image quality and reader experience seen in realworld hip ultrasound, agreement on DDH diagnosis from easily obtained sweeps was only slightly lower than from single images, likely because of the additional step of selecting the best image. AI performed similarly to a nonsubspecialist human reader but was more affected by low-quality images.

show abstract

Section: Discussionmentioning

confidence: 60%

Interobserver Variability of Hip Dysplasia Indices on Sweep Ultrasound for Novices, Experts, and Artificial Intelligence

Ghasseminia

Lim

Concepcion

et al. 2022

Journal of Pediatric Orthopaedics

Self Cite

View full text Add to dashboard Cite

show abstract

“…for hip fractures). Beyond UGRA, use of AI in image interpretation has broader implications across medicine and potentially all of ultrasonography, 23 from screening for developmental dysplasia of the hip 24 to diagnosis of breast cancer. 25 The democratisation of ultrasonography will help ensure that patients have access to the most appropriate interventions, supporting the performance of ultrasound-based interventions by non-experts whilst maintaining relevant clinical standards.…”

Section: Discussionmentioning

confidence: 99%

Assistive artificial intelligence for ultrasound image interpretation in regional anaesthesia: an external validation study

Bowness¹,

Laurent²,

Hernandez³

et al. 2023

British Journal of Anaesthesia

View full text Add to dashboard Cite

“…Hareendranathan et al evaluated the ultrasound image quality according to a 10-point scoring system based on 6 imaging features closely linked to Graf analysis: the presence of the labrum, the os ischium, the midportion of the femoral head, the straightness of the iliac wing, motion, and other artifacts. 31 They used the AI system MEDOHip (MEDO.Ai Inc., Singapore), which calculated the alpha angle and coverage for all slices. MEDOHip uses deep learning to segment the acetabulum, iliac wing, and femoral head and to cover all slices.…”

Section: Discussionmentioning

confidence: 99%

Diagnosis of Developmental Dysplasia of the Hip by Ultrasound Imaging Using Deep Learning

Kinugasa

Inui

Satsuma

et al. 2023

Journal of Pediatric Orthopaedics

View full text Add to dashboard Cite

Background: A timely diagnosis of developmental dysplasia of the hip (DDH) is important for satisfactory clinical outcomes. Ultrasonography is a useful tool for DDH screening; however, it is technically demanding. We hypothesized that deep learning could assist in the diagnosis of DDH. In this study, several deep-learning models were assessed to diagnose DDH on ultrasonograms. This study aimed to evaluate the accuracy of diagnoses made by artificial intelligence (AI) using deep learning on ultrasound images of DDH. Methods: Infants who were up to 6 months old with suspected DDH were included. DDH diagnosis using ultrasonography was performed according to the Graf classification. Data on 60 infants (64 hips) with DDH and 131 healthy infants (262 hips) obtained from 2016 to 2021 were retrospectively reviewed. For deep learning, a MATLAB deep learning toolbox (MathWorks, Natick, MA, US) was used, and 80% of the images were used as training data, with the rest as validation data. Training images were augmented to increase data variation. In addition, 214 ultrasound images were used as test data to evaluate the AI’s accuracy. Pre-trained models (SqueezeNet, MobileNet_v2, and EfficientNet) were used for transfer learning. Model accuracy was evaluated using a confusion matrix. The region of interest of each model was visualized using gradient-weighted class activation mapping (Grad-CAM), occlusion sensitivity, and image LIME. Results: The best scores for accuracy, precision, recall, and F-measure were all 1.0 in each model. In DDH hips, the region of interest for deep learning models was the area lateral to the femoral head, including the labrum and joint capsule. However, for normal hips, the models highlighted the medial and proximal areas where the lower margin of the os ilium and the normal femoral head exist. Conclusions: Ultrasound imaging with deep learning can assess DDH with high accuracy. This system could be refined for a convenient and accurate diagnosis of DDH. Level of Evidence: Level—Ⅳ.

show abstract

Impact of scan quality on AI assessment of hip dysplasia ultrasound

Cited by 12 publications

References 28 publications

Interobserver Variability of Hip Dysplasia Indices on Sweep Ultrasound for Novices, Experts, and Artificial Intelligence

Interobserver Variability of Hip Dysplasia Indices on Sweep Ultrasound for Novices, Experts, and Artificial Intelligence

Assistive artificial intelligence for ultrasound image interpretation in regional anaesthesia: an external validation study

Diagnosis of Developmental Dysplasia of the Hip by Ultrasound Imaging Using Deep Learning

Contact Info

Product

Resources

About