TY - JOUR
T1 - AUTOMATED GLAUCOMA DETECTION USING CENTER SLICE of HIGHER ORDER STATISTICS
AU - Sharma, Rahul
AU - Sircar, Pradip
AU - Pachori, R. B.
AU - Bhandary, Sulatha V.
AU - Acharya, U. Rajendra
PY - 2019/2/1
Y1 - 2019/2/1
N2 - Glaucoma is one of the leading causes of blindness. The raised intraocular pressure is one of the important modifiable risk factor causing glaucomatous optic nerve damage. Glaucomatous optic nerve damage is seen as increase in the cupping of the optic disc and loss of neuroretinal rim. An automated detection system using nonlinear higher order statistics (HOS) based method is used to capture the detailed information present in the fundus image efficiently. The center slice of bispectrum and bicepstrum are applied on fundus images. Various features are extracted from the diagonal of these central slices. In order to reduce the number of features the locality sensitive discriminant analysis (LSDA) data reduction technique method is implemented. The ranked LSDA features are fed to support vector machine (SVM) classifier with various kernels for automated glaucoma detection. The simulation is performed on two databases. The proposed algorithm has yielded classification accuracy of 98.8% and 95% using entire private and public databases, respectively. The proposed technique achieved the highest classification accuracy, hence, confirm the diagnosis of ophthalmologists and can be employed in the community health care centers and hospitals.
AB - Glaucoma is one of the leading causes of blindness. The raised intraocular pressure is one of the important modifiable risk factor causing glaucomatous optic nerve damage. Glaucomatous optic nerve damage is seen as increase in the cupping of the optic disc and loss of neuroretinal rim. An automated detection system using nonlinear higher order statistics (HOS) based method is used to capture the detailed information present in the fundus image efficiently. The center slice of bispectrum and bicepstrum are applied on fundus images. Various features are extracted from the diagonal of these central slices. In order to reduce the number of features the locality sensitive discriminant analysis (LSDA) data reduction technique method is implemented. The ranked LSDA features are fed to support vector machine (SVM) classifier with various kernels for automated glaucoma detection. The simulation is performed on two databases. The proposed algorithm has yielded classification accuracy of 98.8% and 95% using entire private and public databases, respectively. The proposed technique achieved the highest classification accuracy, hence, confirm the diagnosis of ophthalmologists and can be employed in the community health care centers and hospitals.
UR - http://www.scopus.com/inward/record.url?scp=85061540896&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85061540896&partnerID=8YFLogxK
U2 - 10.1142/S0219519419400116
DO - 10.1142/S0219519419400116
M3 - Article
AN - SCOPUS:85061540896
SN - 0219-5194
VL - 19
JO - Journal of Mechanics in Medicine and Biology
JF - Journal of Mechanics in Medicine and Biology
IS - 1
M1 - 1940011
ER -