Skoči na glavni sadržaj

Izvorni znanstveni članak

https://doi.org/10.20532/cit.2017.1003412

Non-invasive Glaucoma Screening Using Ocular Thermal Image Classification

Padmapriya Nammalwar ; Department of Mathematics, SSN College of Engineering, Kalavakkam, India
Venkateswaran Narasimhan ; Department of Electronics and Communication Engineering, SSN College of Engineering, Kalavakkam, India
Toshitha Kannan ; Department of Electronics and Communication Engineering, SSN College of Engineering, Kalavakkam, India
SindhuMadhuri Morapakala ; Department of Electronics and Communication Engineering, SSN College of Engineering, Kalavakkam, India


Puni tekst: engleski pdf 395 Kb

str. 227-236

preuzimanja: 684

citiraj


Sažetak

Ocular thermography is an important, emerging modality in the diagnosis and management of diseases related to eye. It is a non-invasive procedure to evaluate the presence of eye diseases and monitor the response to treatments. In this paper, we propose and evaluate a system designed using infrared thermal image processing that detects glaucoma. Euclidean distance based segmentation technique is used to threshold the IR image to obtain the region of interest, where the manifestation of glaucoma is predominant. Features are extracted using statistical moments from the temperature mapped IR image and Gray Level Co-Occurrence Matrix of the IR image. Two significant attributes, namely the homogeneity and area of region of interest are the inputs to a Support Vector Machine classifier to classify a given input ocular thermal image as a normal or diseased image. In our simulation study, one hundred ocular thermal images with even number of normal and diseased subjects were analysed. The classifier has achieved a maximum accuracy of 96% when homogeneity and area of region of interest are used as attributes, indicating the potential use of proposed method for screening patients even at early stages of glaucoma.

Ključne riječi

thermal imaging; glaucoma; support vector machine; temperature mapping; gray level co-occurrence matrices

Hrčak ID:

188348

URI

https://hrcak.srce.hr/188348

Datum izdavanja:

24.10.2017.

Posjeta: 1.232 *