Shoulder Implant Identification and Categorization with A Fully Implemented Convolution Neural Network

• Author(s): Rajendra Prasad Banavathu ; Prof. M. James Stephen ; Prof. P. V. G. D. Prasad Reddy
• Paper ID: 1704396
• Page: 889-898
• Published Date: 31-05-2023
• Published In: Iconic Research And Engineering Journals
• Publisher: IRE Journals
• e-ISSN: 2456-8880
• Volume/Issue: Volume 6 Issue 11 May-2023

Abstract

Shoulder implants usually need to be changed after a specific time has passed. However, establishing the implant maker or model during this transition may be a complex and error-prone procedure for medical practitioners. This research aims to determine which of four separate implant manufacturers produced each of the 597 X-ray photos of shoulder implants. In order to accomplish this goal, both pre-trained ESA architectures (DenseNet201, DenseNet169, InceptionV3, NasNetLarge, VGG16, VGG19, and Resnet50) and cascading models fed by the YOLOv3 detection algorithm were developed, and the classification performances of these models were compared with one another. The job of the YOLOv3 detection algorithm in the stepped models is to identify the head area of the shoulder implants and provide this region as an input to the ESA designs. This work is performed within the context of the stepped models. In addition, conventional machine learning techniques were integrated with the ensemble learning approach. This integration was analyzed using Fully Implemented Convolution Neural Network (FCNN model) to see how well they performed on the data set. With an accuracy of 84.76 percent, the stepped DenseNet201 model achieved the best classification performance. This rate is greater than the one found in another research that used a comparable data set. The categorization accuracy provided by ensemble models is noticeably lower than that provided by ESA models. Additionally, the classification accuracy achieved by YOLO-assisted cascade models is superior to that achieved by individual ESA models. That is to say, concentrating on the head area of the implant while utilizing the YOLOV3 detection algorithm helped boost the accuracy of the categorization. This methodology will motivate more research into this subject area.

Keywords

YOLO, Shoulder implant, object detection, deep learning, CNN

Citations

IRE Journals:
Rajendra Prasad Banavathu , Prof. M. James Stephen , Prof. P. V. G. D. Prasad Reddy "Shoulder Implant Identification and Categorization with A Fully Implemented Convolution Neural Network" Iconic Research And Engineering Journals Volume 6 Issue 11 2023 Page 889-898

IEEE:
Rajendra Prasad Banavathu , Prof. M. James Stephen , Prof. P. V. G. D. Prasad Reddy "Shoulder Implant Identification and Categorization with A Fully Implemented Convolution Neural Network" Iconic Research And Engineering Journals, 6(11)