Back

Novel scotoma detection method using time required for fixation to the random targets

Takahashi, N.; Saeki, S.; Kawahara, M.; Aman, H.; Nakano, E.; Mori, Y.; Miyake, M.; Tamura, H.; Tsujikawa, A.

2021-06-05 ophthalmology
10.1101/2021.06.03.21258101 medRxiv
Show abstract

We developed a novel scotoma detection system using time required for fixation to the random targets, or the" eye-guided scotoma detection method ". In order to verify the" eye-guided scotoma detection method ", we measured 78 eyes of 40 subjects, and examined the measurement results in comparison with the results of measurement by Humphrey perimetry. The results were as follows: (1) Mariotte scotomas were detected in 100% of the eyes tested; (2) The false-negative rate (the percentage of cases where a scotoma was evaluated as a non-scotoma) was less than 10%; (3) The positive point distribution in the low-sensitivity eyes was well matched. These findings suggested that the novel scotoma detection method in the current study will pave the way for the realization of mass screening to detect pathological scotoma earlier. Author summaryConventional perimeters, such as the Goldmann perimeter and Humphrey perimeter, require experienced examiners and space occupying. With either perimeter, subjects eye movements need to be strictly fixed to the fixation target of the device. Other perimeters can monitor fixation and automatically measure the visual field. With the eye-guided scotoma detection method proposed in the current study, subjects feel less burdened since they do not have to fixate on the fixation target of the device and can move their eyes freely. Subjects simply respond to visual targets on the display; then, scotomas can be automatically detected. The novel method yields highly accurate scotoma detection through an algorithm that separates scotomas from non-scotomas.

Matching journals

The top 2 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.