June 2023
Volume 64, Issue 8
Open Access
ARVO Annual Meeting Abstract  |   June 2023
Understanding Convolutional Neural Network model decision for Glaucoma detection using Gradient class activation maps based on Compass color fundus images
Author Affiliations & Notes
  • Silvia Gazzina
    Centervue, Italy
  • CHIARA RUI
    Centervue, Italy
  • Dario Romano
    Eye Clinic, Università degli Studi di Milano, Italy
  • Benedetta Colizzi
    Eye Clinic, Università degli Studi di Milano, Italy
  • Paolo Fogagnolo
    Eye Clinic, Università degli Studi di Milano, Italy
  • Luca Mario Rossetti
    Eye Clinic, Università degli Studi di Milano, Italy
  • Footnotes
    Commercial Relationships   Silvia Gazzina Centervue, Code E (Employment); CHIARA RUI Centervue, Code E (Employment); Dario Romano None; Benedetta Colizzi None; Paolo Fogagnolo Centervue, Code C (Consultant/Contractor); Luca Rossetti Centervue, Code C (Consultant/Contractor)
  • Footnotes
    Support  None
Investigative Ophthalmology & Visual Science June 2023, Vol.64, 377. doi:
  • Views
  • Share
  • Tools
    • Alerts
      ×
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Silvia Gazzina, CHIARA RUI, Dario Romano, Benedetta Colizzi, Paolo Fogagnolo, Luca Mario Rossetti; Understanding Convolutional Neural Network model decision for Glaucoma detection using Gradient class activation maps based on Compass color fundus images. Invest. Ophthalmol. Vis. Sci. 2023;64(8):377.

      Download citation file:


      © ARVO (1962-2015); The Authors (2016-present)

      ×
  • Supplements
Abstract

Purpose : To compare Gradient class activation maps (Grad-CAM) heatmaps of three Convolutional Neural Networks (CNNs) in localizing glaucomatous optic neuropathy (GON) and to evaluate their clinical usefulness

Methods : The data used for this project consisted of 193 color retinal images including the Optic Nerve Head (ONH) collected with the Compass fundus perimeter (Centervue, Italy), for a total of 101 images labelled as healthy (NRM) and 92 as glaucoma (GLC). The data were processed to obtain different crop sizes as input for three CNNs: Square Crop (SC), Halfmoon Crop (HC) and Fovea-Extended Crop (FEC) models, respectively using a 200x200, 700x300 and 700x500 pixel input crop. Grad-CAM Heatmaps (GCH) were generated for the three CNNs (example shown in Fig1a). Last convolutional layer (CL) and average GCH were analyzed for all CNNs to understand the decision-making process. A qualitative approach was adopted to classify GCH into 4 pre-defined categories. Fig1b provides a visual explanation for each category. The percentage of activation (PA) of each category was compared among CNNs

Results : The results are reported in Fig2. For images predicted as GLC by the CNNs, GCH should be active on image portions involving ONH, which are clinically meaningful for GON. This is supported by the high PA of GCH in the category involving ONH for the HC model. On the other hand, for images predicted as NRM, GCH should not highlight any image area. However, the results show a low percentage of GCH in the category “no activation” and a high percentage in the category involving crop edges and ONH

Conclusions : The results indicate that GCH are reasonable for images predicted as GLC, whereas they may not be clinically useful for images predicted as NRM. HC model shows the higher PA on image portions involving ONH, suggesting that this is the most clinically meaningful model. In conclusion, GCH of HC model could be useful in a glaucoma-care clinical setting to better understand the CNN outcome

This abstract was presented at the 2023 ARVO Annual Meeting, held in New Orleans, LA, April 23-27, 2023.

 

Fig1a: example of GCH for a test set image derived from HC model. From left to right, GCH are displayed for each CL. The rightmost image is the averaged GCH for the last four CL; Fig1b: GCH classification categories

Fig1a: example of GCH for a test set image derived from HC model. From left to right, GCH are displayed for each CL. The rightmost image is the averaged GCH for the last four CL; Fig1b: GCH classification categories

 

Fig2: PA of categories in GCH for all predictions. The blue and orange bars indicate the PA of categories for last CL and average GCH of the three CNNs respectively

Fig2: PA of categories in GCH for all predictions. The blue and orange bars indicate the PA of categories for last CL and average GCH of the three CNNs respectively

×
×

This PDF is available to Subscribers Only

Sign in or purchase a subscription to access this content. ×

You must be signed into an individual account to use this feature.

×