TY  - JOUR
T1  - Automatic Assessment for the Detection of Knee Effusion using Magnetic Resource Imaging
AU - Yousuf Bhat, Aamir AU - Suhasini, A. 
JO  - Asian Journal of Information Technology
VL  - 19
IS  - 3
SP  - 70
EP  - 81
PY  - 2020
DA  - 2001/08/19
SN  - 1682-3915
DO  - ajit.2020.70.81
UR  - https://makhillpublications.co/view-article.php?doi=ajit.2020.70.81
KW  - Osteoarthritis (OA)
KW  -Knee effusion
KW  -2-D gabor filter
KW  -random forest
KW  -multi-layer BPNN
KW  -adaboost SVM
AB  - Effusion of the knee joint is possibly related to
osteoarthritis erupt and is a significant marker of remedial
reaction. The investigation is planned for creating and
approving a computerized framework dependent on MR
imaging for the measurement of joint effusion. The
occurrence of knee effusion requires an extensive
differential determination and an orderly symptomatic
approach. Yearning of the knee effusion is a
fundamentally demonstrative and restorative intercession
in numerous rheumatologic diseases. The clinical
investigation has traditionally included tests counting the
patella tap. The precision of these tests for identifying the
effusion and measure is not well set up. MR imaging is
considered superior for recognizable proof and evaluation
of knee effusion. The amount of effusion present in the
joint was recorded and MRI criteria for the detection of
knee effusion were assessed. The fat cushion division
sign was the foremost exact marker of liquid as little
as 1-2 mLwas recognized. Axial view of MRI images was
used in accessing the knee effusion. The classifier was
superior both in terms of time efficiency and classification
performance to classifier regularly used on the basis of
iterative learning. In this paper we have used two features
namely watershed Segmentation and 2-D Gabor Filter.
The extracted features from MRI image are given to the
classifiers namely Random Forest, Multi Linear BPNN
and Adaboost SVM. The random forest classifier was
good when comparing with the other two classifier and
achieves the good accuracy rate of 92.12%. Finally, the
classifier was prevalent both in time adequacy and order
execution to the routinely used classifiers dependent on
iterative learning.
ER  -