Eventos Anais de eventos
COBEM 2021
26th International Congress of Mechanical Engineering
PILOT: ARTIFICIAL INTELLIGENCE APPLIED TO THE IDENTIFICATION OF BONE CHANGES IN CANINE PELVIC RADIOGRAPHIES
Submission Author:
Barbara Emmanuelle Sanches Silva , MG , Brazil
Co-Authors:
Barbara Emmanuelle Sanches Silva, Rudolf Huebner, Anelise Nepomuceno, Carolina Costa Cardoso
Presenter: Barbara Emmanuelle Sanches Silva
doi://10.26678/ABCM.COBEM2021.COB2021-1052
Abstract
Artificial intelligence has been applied to the health sciences as a new method of metrology and instrumentation as it uses a database to create statistical reference standards related to this database. In veterinary medicine, the low structure and the great diversity of data results in a still incipient use of this technology. The application of artificial intelligence as an auxiliary diagnostic method in images is already used in human medicine with high precision and accuracy to aid decision making, especially, in the identification of changes in physiological patterns in X-rays. This work proposes a comparison of application of a convolutional neural network (CCN) and the neural network Multi-layer Perceptron for identification of bone changes of different causes in ventrodorsal x-rays of dogs of different breeds and sizes. Images with normal radiological patterns and with different radiological alterations were collected from different radiology clinics to create the database, dividing the data between normal and not normal patterns. Database expansion techniques were used to avoid overfitting, resized and the images were divided into training (90% and 83%) and testing (10% and 17%). The performances of the neural networks are compared to the radiological reports issued by human specialists. The results of sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV) and F1-Score are obtained. The results aim to show that, even for data with high variability, the application of neural networks as an auxiliary method of diagnosis is a valuable resource in veterinary medicine and enlargement of the dataset is recommended.
Keywords
convolutional neural network, Artificial Intelligence, radiology, veterinary medicine
