Fundus images with glaucoma and glaucoma suspect through CLAHE with reverse polar transformation, GradCAM++ heatmap of ConvNeXt-S model trained on polar with “donut method”, attention map of DeiT- S model trained on polar with “pie method” with gradient of predicted class, GradCAM++ heatmap overlay on reverse polar and attention map overlay on reverse polar.

<p>Fundus images with glaucoma and glaucoma suspect through CLAHE with reverse polar transformation, GradCAM++ heatmap of ConvNeXt-S model trained on polar with “donut method”, attention map of DeiT- S model trained on polar with “pie method” with gradient of predicted class, GradCAM++ heatmap...

Full description

Saved in:

Bibliographic Details
Main Author:	Sirikorn Sangchocanonta (20917797) (author)
Other Authors:	Pakinee Pooprasert (14886029) (author), Nichapa Lerthirunvibul (20917800) (author), Kanyarak Patchimnan (20917803) (author), Phongphan Phienphanich (14886035) (author), Adirek Munthuli (14886026) (author), Sujittra Puangarom (20917806) (author), Rath Itthipanichpong (20495532) (author), Kitiya Ratanawongphaibul (20495523) (author), Sunee Chansangpetch (9427653) (author), Anita Manassakorn (9992809) (author), Visanee Tantisevi (9992812) (author), Prin Rojanapongpun (9992815) (author), Charturong Tantibundhit (14886047) (author)
Published:	2025
Subjects:	Medicine Cell Biology Genetics Physiology Space Science Biological Sciences not elsewhere classified Mathematical Sciences not elsewhere classified Information Systems not elsewhere classified reduced computational demands pie method ," meticulously annotated glaucutu employing polar transformation convolutional neural network clinically significant areas augmentation method aligned 45 %, respectively 19 %, respectively 05 %, respectively 63 &# 177 35 &# 177 32 &# 177 23 &# 177 22 &# 177 preprocessing time decreased efficient image transformers irreversible vision impairment xlink "> glaucoma 09 &# 177 deit &# 8217 elevated performance levels 70 %, 84 53 %, 84 diagnosing glaucoma increased 45 %, 68 39 %, 92 &# 177 39 %, vision transformers &# 8220 69 %, 54 %, time challenges 09 %. vision transformer deit facilitated deit ), achieved 84 glaucoma suspects glaucoma screening unanimous agreement thereby enhancing temporal regions standard deviation sensitivity increased precise localization overall effectiveness nasal regions majority agreement leading cause isnt rule initially lower fundus imaging early detection disc rim critical need average area attention maps 24 %.
Tags:	Add Tag No Tags, Be the first to tag this record!

Similar Items