ApproxNet: Content and Contention-Aware Video Object Classification System for Embedded Clients

Videos take a lot of time to transport over the network, hence running analytics on the live video on embedded or mobile devices has become an important system driver. Considering such devices, e.g., surveillance cameras or AR/VR gadgets, are resource constrained, although there has been significant work in creating lightweight deep neural networks (DNNs) for such clients, none of these can adapt to changing runtime conditions, e.g., changes in resource availability on the device, the content characteristics, or requirements from the user. In this article, we introduce ApproxNet, a video object classification system for embedded or mobile clients. It enables novel dynamic approximation techniques to achieve desired inference latency and accuracy trade-off under changing runtime conditions. It achieves this by enabling two approximation knobs within a single DNN model rather than creating and maintaining an ensemble of models, e.g., MCDNN [MobiSys-16]. We show that ApproxNet can adapt seamlessly at runtime to these changes, provides low and stable latency for the image and video frame classification problems, and shows the improvement in accuracy and latency over ResNet [CVPR-16], MCDNN [MobiSys-16], MobileNets [Google-17], NestDNN [MobiCom-18], and MSDNet [ICLR-18].

Enhancing the radiological classification system from the distal femur to the proximal tibia

Introduction The inner diaphyseal diameter of the distal femur, at 20 cm from the lateral joint line, is the strongest risk factor for predicting aseptic loosening in total knee arthroplasty using rotating hinge prosthesis. In this context, the Citak classification has been introduced presenting three different types of the distal femur anatomy. The aim of the study is to develop a novel classification system for the proximal tibia. Materials and Methods Two-hundred patients with standard knee antero-posterior radiographs were included in this study. We measured the inner diameter of the tibia 16 cm distally from the tibial plateau and 3 cm distally from the tibial spine. The ratio between these two measurements was applied as the novel index ratio. Results According to the 25th and 75th percentiles, three groups can be clustered for each gender. A higher distribution of the type B pattern was found in female and male patients. However, type A with a narrow inner diaphyseal diameter was less common in female patients The median intra-observer reliability for rater 1 was 0.997. The inter-observer reliability was high (ICC 0.998). There was a moderate correlation between the AP diameter and height (r = 0,568); a low correlation between the AP diameter and weight (r = 0.376). The novel index shows no significant correlation between the index ratio and height (r = 0.082), weight (r = 0.014) or BMI (r= - 0.038). The novel index shows no statistically significant correlation between the index ratio and height (r = 0.082) or weight (r = 0.014) or BMI (r= - 0.038). Conclusion The novel classification presents three different types of tibia for each gender: type C has a wider inner diaphyseal diameter compared to type A with a narrow inner diaphyseal diameter. Type B has the widest distribution among the subjects.

The Influence of Regional Differences on the Reliability of the AO Spine Sacral Injury Classification System

Study Design Global cross-sectional survey. Objective To explore the influence of geographic region on the AO Spine Sacral Classification System. Methods A total of 158 AO Spine and AO Trauma members from 6 AO world regions (Africa, Asia, Europe, Latin and South America, Middle East, and North America) participated in a live webinar to assess the reliability, reproducibility, and accuracy of classifying sacral fractures using the AO Spine Sacral Classification System. This evaluation was performed with 26 cases presented in randomized order on 2 occasions 3 weeks apart. Results A total of 8320 case assessments were performed. All regions demonstrated excellent intraobserver reproducibility for fracture morphology. Respondents from Europe (k = .80) and North America (k = .86) achieved excellent reproducibility for fracture subtype while respondents from all other regions displayed substantial reproducibility. All regions demonstrated at minimum substantial interobserver reliability for fracture morphology and subtype. Each region demonstrated >90% accuracy in classifying fracture morphology and >80% accuracy in fracture subtype compared to the gold standard. Type C morphology (p2 = .0000) and A3 (p1 = .0280), B2 (p1 = .0015), C0 (p1 = .0085), and C2 (p1 =.0016, p2 =.0000) subtypes showed significant regional disparity in classification accuracy (p1 = Assessment 1, p2 = Assessment 2). Respondents from Asia (except in A3) and the combined group of North, Latin, and South America had accuracy percentages below the combined mean, whereas respondents from Europe consistently scored above the mean. Conclusions In a global validation study of the AO Spine Sacral Classification System, substantial reliability of both fracture morphology and subtype classification was found across all geographic regions.