Forecasting Future Humphrey Visual Fields Using Deep Learning

Abstract

Purpose: To determine if deep learning networks could be trained to forecast a future 24-2 Humphrey Visual Field (HVF).Design: Retrospective database study.Participants: All patients who obtained a HVF 24-2 at the University of Washington.Methods: All datapoints from consecutive 24-2 HVFs from 1998 to 2018 were extracted from a University of Washington database. Ten-fold cross validation with a held out test set was used to develop the three main phases of model development: model architecture selection, dataset combination selection, and time-interval model training with transfer learning, to train a deep learning artificial neural network capable of generating a point-wise visual field prediction. Main outcome measures: Mean absolute error (MAE) and difference in Mean Deviation (MD) between predicted and actual future HVF.Results: More than 1.7 million perimetry points were extracted to the hundredth decibel from 32,443 24-2 HVFs. The best performing model with 20 million trainable parameters, CascadeNet-5, was selected. The overall MAE for the test set was 2.47 dB (95% CI: 2.45 dB to 2.48 dB). The 100 fully trained models were able to successfully predict progressive field loss in glaucomatous eyes up to 5.5 years in the future with a correlation of 0.92 between the MD of predicted and actual future HVF (p < 2.2 x 10 -16 ) and an average difference of 0.41 dB.Conclusions: Using unfiltered real-world datasets, deep learning networks show an impressive ability to not only learn spatio-temporal HVF changes but also to generate predictions for future HVFs up to 5.5 years, given only a single HVF.

A Review of Learning Effect in Perimetry

“…The skill to quickly predict future glaucomatous progression may prevent pointless functional loss that can occur with the contemporary practice of multiple confirmatory VF tests. In the near future, after incorporation of clinical data such as IOP, medication and surgical history, this model may give a hand in clinical decision-making and allow development of a personalized treatment regimen for each patient [23].…”

“…As a final point, Wen and colleagues in recent study (2019), using unfiltered real-world datasets of deep learning networks, show the ability to not only learn spatio-temporal Humphrey Visual Fields (HVF) changes but also to generate predictions for future HVFs up to 5.5 years, given only a single HVF [23].…”

“…The 1st visit filtered outcome can be compared with the 5th visit raw sensitivity chart, where the matching locations of the visual field are shaded yellow presuppose a continuous rate of VF worsening and fail to concentrate on varying rates of glaucomatous deterioration over time. Wen and colleagues (2019) by using deep learning models trained on the temporal history for a large group of patients, and they claim that produced singleHumphrey VFs and predicted point-wise VF printouts at half-year time intervals, up to five years later[23], In an ideal world, forecasted VFs would guide clinical decision-making and disease management.…”

Accurate prediction of glaucoma from colour fundus images with a convolutional neural network that relies on active and transfer learning

“…), VF prediction (Wen et al. ), and automated glaucoma detection using small data sets (Matsopoulos et al. ; Asaoka et al.…”

How Artificial Intelligence Can Transform Randomized Controlled Trials

“…As a proof of concept, DL models have been able to predict what Humphrey visual field (HVF) would appear in up to 5.5 years from a single baseline HVF while ingesting clinical metadata. 17 The algorithm will need to be validated in independent populations but provides preliminary data that AI models could be used to predict disease progressions in synthetic controls. The prediction models would ingest the sum total of clinical, genetic, and imaging data to generate future progression of disease for each trial participant.…”