Retinal Specialist versus Artificial Intelligence Detection of Retinal Fluid from OCT: Age-Related Eye Disease Study 2: 10-Year Follow-On Study

PURPOSE

To evaluate the performance of retinal specialists in detecting retinal fluid presence in spectral domain OCT (SD-OCT) scans from eyes with age-related macular degeneration (AMD) and compare performance with an artificial intelligence algorithm.

DESIGN

Prospective comparison of retinal fluid grades from human retinal specialists and the Notal OCT Analyzer (NOA) on SD-OCT scans from 2 common devices.

PARTICIPANTS

A total of 1127 eyes of 651 Age-Related Eye Disease Study 2 10-year Follow-On Study (AREDS2-10Y) participants with SD-OCT scans graded by reading center graders (as the ground truth).

METHODS

The AREDS2-10Y investigators graded each SD-OCT scan for the presence/absence of intraretinal and subretinal fluid. Separately, the same scans were graded by the NOA.

MAIN OUTCOME MEASURES

Accuracy (primary), sensitivity, specificity, precision, and F1-score.

RESULTS

Of the 1127 eyes, retinal fluid was present in 32.8%. For detecting retinal fluid, the investigators had an accuracy of 0.805 (95% confidence interval [CI], 0.780-0.828), a sensitivity of 0.468 (95% CI, 0.416-0.520), a specificity of 0.970 (95% CI, 0.955-0.981). The NOA metrics were 0.851 (95% CI, 0.829-0.871), 0.822 (95% CI, 0.779-0.859), 0.865 (95% CI, 0.839-0.889), respectively. For detecting intraretinal fluid, the investigator metrics were 0.815 (95% CI, 0.792-0.837), 0.403 (95% CI, 0.349-0.459), and 0.978 (95% CI, 0.966-0.987); the NOA metrics were 0.877 (95% CI, 0.857-0.896), 0.763 (95% CI, 0.713-0.808), and 0.922 (95% CI, 0.902-0.940), respectively. For detecting subretinal fluid, the investigator metrics were 0.946 (95% CI, 0.931-0.958), 0.583 (95% CI, 0.471-0.690), and 0.973 (95% CI, 0.962-0.982); the NOA metrics were 0.863 (95% CI, 0.842-0.882), 0.940 (95% CI, 0.867-0.980), and 0.857 (95% CI, 0.835-0.877), respectively.

CONCLUSIONS

In this large and challenging sample of SD-OCT scans obtained with 2 common devices, retinal specialists had imperfect accuracy and low sensitivity in detecting retinal fluid. This was particularly true for intraretinal fluid and difficult cases (with lower fluid volumes appearing on fewer B-scans). Artificial intelligence-based detection achieved a higher level of accuracy. This software tool could assist physicians in detecting retinal fluid, which is important for diagnostic, re-treatment, and prognostic tasks.

Signal deconvolution based expression-detection and background adjustment for microarray data

Background adjustment is an essential stage in analyzing DNA microarrays. Discriminating expressed genes from unexpressed ones (expression detection), and estimating the expression levels of weakly expressed genes, critically depend on accurate treatment of the background intensity. Current methods for background adjustment either do not deal with nonspecific hybridization or strongly depend on the reliability of control probes. Existing model-based methods have limited accuracy. A new platform-independent background adjustment algorithm is presented. The algorithm relies on the deconvoluted experimental signal distribution for evaluating the expression probability and adjusting the background of each probe. Considering expression detection, it is shown, for two-channels cDNA arrays and for the Affymetrix GeneChip platform, that the algorithm performs at least as good or better than control-probes-based algorithms. For the Affymetrix GeneChip arrays, it is further shown that the algorithm outperforms the robust multiarray (RMA) expression measure in estimating genomewide expression levels.

Large-scale unrestricted identification of post-translation modifications using tandem mass spectrometry

TwinPeaks, a close variant of the SEQUEST protein identification algorithm, is capable of unrestricted, large-scale, identification of post-translation modifications (PTMs). TwinPeaks is applied on a sample of 100441 tandem mass spectra from the HUPO Plasma Proteome Project data set, with full non-redundant human as a reference protein database. With a 3.5% error rate, TwinPeaks identifies a collection of 539 spectra that were not identified by the usual PTM-restricted identification algorithm. At this error rate, TwinPeaks increases the rate of spectra identifications by at least 17.6%, making unrestricted PTM identification an integral part of proteomics.

Evidence for abundant transcription of non-coding regions in the Saccharomyces cerevisiae genome

Background

Recent studies in a growing number of organisms have yielded accumulating evidence that a significant portion of the non-coding region in the genome is transcribed. We address this issue in the yeast Saccharomyces cerevisiae.

Results

Taking into account the absence of a significantly large yeast EST database, we use microarray expression data collected for genomic regions erroneously believed to be coding to study the expression pattern of non-coding regions in the Saccharomyces cerevisiae genome. We find that at least 164 out of 589 (28%) such regions are expressed under specific biological conditions. In particular, looking at the probes that are located opposing other known genes at the same genomic locus, we find that 88 out of 341 (26%) of these genes support antisense transcription. The expression patterns of these antisense genes are positively correlated. We validate these results using RT-PCR on a sample of 6 non-coding transcripts.

Conclusion

1. The yeast genome is transcribed on a scale larger than previously assumed. 2. Correlated transcription of antisense genes is abundant in the yeast genome. 3. Antisense genes in yeast are non-coding.

Intensity-based statistical scorer for tandem mass spectrometry

We describe a new statistical scorer for tandem mass spectrometry. The scorer is based on the probability that fragments with given chemical properties create measured intensity levels in the experimental spectrum. The scorer's parameters are computed using a fully automated procedure. Benchmarking the new scorer on a large set of experimental spectra, we show that it performs significantly better than the widely used cross-correlation scoring algorithm of Eng et al. (Eng, J. K; McKormack, A. L.; Yates, J. R. J. Am. Soc. Mass Spectrom. 1994, 5, 976-989.).