TY - JOUR
T1 - Erratum
T2 - Unpaired intra-operative OCT (iOCT) video super-resolution with contrastive learning: erratum
AU - Komninos, Charalampos
AU - Pissas, Theodoros
AU - Flores, Blanca
AU - Bloch, Edward
AU - Vercauteren, Tom
AU - Ourselin, Sébastien
AU - Da Cruz, Lyndon
AU - Bergeles, Christos
N1 - Publisher Copyright:
Journal © 2025.
PY - 2025/2/1
Y1 - 2025/2/1
N2 - Errata are presented to amend the values presented in results section and corresponding figures in our published manuscript [Biomed. Opt. Express 15, 772 (2024)]. These corrections do not alter the main findings of our publication. Published by Optica Publishing Group under the terms of the Creative Commons Attribution 4.0 License. Further distribution of this work must maintain attribution to the author(s) and the published article’s title, journal citation, and DOI. Due to incosistencies in the number of iOCT image sequences and the number of frames per sequence utilised for evaluation across different approaches, we provide corrections to the tables and figures in [1] as follows, with the corrected method(s) indicated in parentheses: Table 1 ([2], Ours), Fig. 8 (4th column), Table 2 (w/o L
Contr(λ
3 = 1), (λ
3 = 5), (λ
3 = 10), w/o L
GAN, Ours), Table 3 (ours-2frames, Ours), Fig. 10 (FID and |∆GCF| graphs) and Table 4 (NLM, NLM(σ̃), Ours). (Figure Presented) Furthermore, we correct related phrases in the text as follows: (Figure Presented) Section 3.1, 2nd paragraph, 2nd sentence is corrected as follows: Our unpaired video super-resolution approach demonstrates superior performance compared to all the other iOCT super-resolution methods, ranking first in five out of nine metrics. Section 3.1, 4th paragraph, 1st and 2nd sentences are corrected as follows: Moreover, recognising that perceptual metrics may not be able to capture and assess low-level characteristics, we provide an analysis of the |∆GCF| and |∆FNE| values, where our method achieves the second lowest (best) values. Section 3.3, 2nd paragraph, 2nd sentence is corrected as follows: Removing the (w/o L
Contr, λ
3 = 1) contrastive term and keeping the perceptual term’s weight in its default value (λ
3 = 1) results in worse super-resolution results, as reported in Table 2. Section 3.3, 3rd paragraph, 1st and 2nd sentences are corrected as follows: Furthermore, removing the contrastive loss term and adjusting the weights for the perceptual loss, (w/o L
Contr, λ
3 = 5 and w/o L
Contr, λ
3 = 10), improves performance but still results in slightly worsen quality metrics. Our method outperforms the w/o L
Contr approaches in most of the metrics indicating that the perceptual quality and noise levels of the preOCT are better captured by the model when using our complete loss function ("ours"). Regarding Table 1, after the corrections, our approach (Ours) loses its superiority in |∆GCF| but remains best in most of the metrics. Regarding the comparison with the w/o L
Contr variations in Table 2, after the corrections our (Table Presented) approach (Ours) loses its superiority in |∆GCF| and Coverage(C) metrics while achieves best value in Density (D). As shown in Table 2, it remains superior in most of the metrics. Particularly, (Table Presented) it outperforms the w/o L
Contr, λ
3 = 1 and λ
3 = 10 in 5 out of 6 metrics and the w/o L
Contr, λ
3 = 5 in 4 out of 6 metrics. Similarly, concerning the 9 metrics, our method outperforms the w/o L
Contr, λ
3 = 1 and λ
3 = 10 in 7 out of 9 metrics and the w/o L
Contr, λ
3 = 5 in 5 out of 9 metrics. In Table 2, the reported values for our approach differ from those in other tables because we use a checkpoint obtained at a later stage of the training. The results reported in all the remaining tables, as well as the qualitative analysis and figures, are derived from the same checkpoint. The reported corrections do not change the information presented in the abstract, introduction, methods and discussion sections of the original paper, and they do not alter the main conclusions of our manuscript. Therefore, limited corrections are required in the main text.
AB - Errata are presented to amend the values presented in results section and corresponding figures in our published manuscript [Biomed. Opt. Express 15, 772 (2024)]. These corrections do not alter the main findings of our publication. Published by Optica Publishing Group under the terms of the Creative Commons Attribution 4.0 License. Further distribution of this work must maintain attribution to the author(s) and the published article’s title, journal citation, and DOI. Due to incosistencies in the number of iOCT image sequences and the number of frames per sequence utilised for evaluation across different approaches, we provide corrections to the tables and figures in [1] as follows, with the corrected method(s) indicated in parentheses: Table 1 ([2], Ours), Fig. 8 (4th column), Table 2 (w/o L
Contr(λ
3 = 1), (λ
3 = 5), (λ
3 = 10), w/o L
GAN, Ours), Table 3 (ours-2frames, Ours), Fig. 10 (FID and |∆GCF| graphs) and Table 4 (NLM, NLM(σ̃), Ours). (Figure Presented) Furthermore, we correct related phrases in the text as follows: (Figure Presented) Section 3.1, 2nd paragraph, 2nd sentence is corrected as follows: Our unpaired video super-resolution approach demonstrates superior performance compared to all the other iOCT super-resolution methods, ranking first in five out of nine metrics. Section 3.1, 4th paragraph, 1st and 2nd sentences are corrected as follows: Moreover, recognising that perceptual metrics may not be able to capture and assess low-level characteristics, we provide an analysis of the |∆GCF| and |∆FNE| values, where our method achieves the second lowest (best) values. Section 3.3, 2nd paragraph, 2nd sentence is corrected as follows: Removing the (w/o L
Contr, λ
3 = 1) contrastive term and keeping the perceptual term’s weight in its default value (λ
3 = 1) results in worse super-resolution results, as reported in Table 2. Section 3.3, 3rd paragraph, 1st and 2nd sentences are corrected as follows: Furthermore, removing the contrastive loss term and adjusting the weights for the perceptual loss, (w/o L
Contr, λ
3 = 5 and w/o L
Contr, λ
3 = 10), improves performance but still results in slightly worsen quality metrics. Our method outperforms the w/o L
Contr approaches in most of the metrics indicating that the perceptual quality and noise levels of the preOCT are better captured by the model when using our complete loss function ("ours"). Regarding Table 1, after the corrections, our approach (Ours) loses its superiority in |∆GCF| but remains best in most of the metrics. Regarding the comparison with the w/o L
Contr variations in Table 2, after the corrections our (Table Presented) approach (Ours) loses its superiority in |∆GCF| and Coverage(C) metrics while achieves best value in Density (D). As shown in Table 2, it remains superior in most of the metrics. Particularly, (Table Presented) it outperforms the w/o L
Contr, λ
3 = 1 and λ
3 = 10 in 5 out of 6 metrics and the w/o L
Contr, λ
3 = 5 in 4 out of 6 metrics. Similarly, concerning the 9 metrics, our method outperforms the w/o L
Contr, λ
3 = 1 and λ
3 = 10 in 7 out of 9 metrics and the w/o L
Contr, λ
3 = 5 in 5 out of 9 metrics. In Table 2, the reported values for our approach differ from those in other tables because we use a checkpoint obtained at a later stage of the training. The results reported in all the remaining tables, as well as the qualitative analysis and figures, are derived from the same checkpoint. The reported corrections do not change the information presented in the abstract, introduction, methods and discussion sections of the original paper, and they do not alter the main conclusions of our manuscript. Therefore, limited corrections are required in the main text.
UR - http://www.scopus.com/inward/record.url?scp=85216898085&partnerID=8YFLogxK
U2 - 10.1364/BOE.555423
DO - 10.1364/BOE.555423
M3 - Article
C2 - 39958856
SN - 2156-7085
VL - 16
SP - 790
EP - 792
JO - Biomedical Optics Express
JF - Biomedical Optics Express
IS - 2
ER -