Optical Attenuation Coefficient Optimization Algorithm for Deep Tissue Signals in Optical Coherence Tomography Based on Kalman Filter

Round 1

Reviewer 1 Report

The author describes a Kalman filter-based optical attenuation coefficient estimation algorithm that effectively reduces the noise floor effect. It is recommended to improve the manuscript based on the following opinions so that readers can understand and improve the paper.

 1.   The depth decomposition method proposed by Vermeer et al. approximates infinite data using finite data in line 18 of page 2. However, the equations for finite and infinite data shown in the document are the same. The author should provide a clear and exact expression about it.

2.  In page 4, the vertical axis label in Figure 1(b) should be changed to "OAC (MM-1)"

3.  In Eq.(3), the author argues that classical Kalman filters have a control vector 'u' excluded from the "depth-phase" model. It is necessary to explain the reasons more clearly, such as the conditions under which it can be ignored.

4. As shown in Figures 5 and 6, noise resistance to additive and multiplication noise was evaluated in homogeneous media. Figures 7 and 8 show the applicability of the proposed algorithm, but noise tolerance is not evaluated for multiple layers. It is desirable to include an experiment in which a sample with a constant optical attenuation coefficient with a plurality of layers is constructed using a plurality of microscopic cover glass. In Figure 8, it is necessary to change the label to 'Original OAC' instead of 'Optimized OAC'.

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 2 Report

This paper reports the application of Kalman filtering to OAC (optical attenuation coefficient calculation) in OCT (optical coherence tomography).  It is a follow-up work of the authors' previous studies [ref. 23].  In general, the paper is well-written and the results should be of interests to the OCT community.  There are a few issues that the authors should address:

1.  It would help the readers if the authors comment on the quality of the OCT images in this work with those by other groups (using different approaches).

2. In this paper the term OCT-AI was used with "AI" meant for attenuation imaging.  Now a days, "AI" is generally understood as the abbreviation for "artificial intelligence".  The term could be misleading.

In Sec. 3.3, "in vivo experiments", the description and discussion need be clarified:

a) Define clearly what high-error OAC images meant.  Are the errors artificially introduced?

b)  The image quality should be quantitatively accessed, e.g., using parameters such as signal-to-noise ration (SNR), contrast-to-noise ratio (CNR), equivalent number of looks (ENL) and cross-correlation value (XCOR), etc.  

c) There are five subjects. Figures 7 and 8 only show four.  In Fig. 8, two "Optimized OAC" are shown.  Please explain.  Besides, the authors should inform whether differences are observed from the different subjects or not. Are these typical prototypes that are also reported by other references?  

 

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 3 Report

In this manuscript, the authors proposed an enhanced method for the calculation of optical attenuation coefficient by applying a Kalman filter. The authors explained well their work, but there are some parts that need to be described in more detail. With our comments, we hope that the manuscript would be more fruitful.

 

1.      Authors mentioned that effectively removing the noise floor and protecting the tail-end weak signal were important for calculating the OAC of a deep tissue. In Figure 3, to protect the tail-end weak signal, the Kalman filter was applied and to remove the noise floor b, the reference arm spectrum was utilized. Are there any other comparable methods for removing the noise floor?

 

2.      In Figure 4(a), the OCT signal obtained by applying the Kalman filter looks like a low-pass filtered signal. How does this Kalman filter differ from the traditional denoising algorithms mentioned in the introduction part? Can you compare the OAC calculated by the Kalman filter with the one calculated with a conventionally denoised OCT signal, not the “high error OAC” where only the noise floor was subtracted from the OCT signal in Figure 2?

 

3.      In Figure 7(c)-(e), it seems like that there is no significant difference in the OAC value above the RPE layer, but there is large difference below it. Is there any reason for that? When there is a strong reflective layer in the middle of the interesting region, would it give great error in calculation?

 

4.      In Figure 8, there are differences in the OAC values obtained with different methods. Is the difference large enough to serve as a criterion for clearly distinguishing different tissues as mentioned in the introduction part? Are there any examples of tissues that have been distinguished by these minute OAC differences?

 

5.      The conclusion part needs more information or contents.

 

6.      Minor errors;

1)     Second paragraph in page 2, what is the difference between the “finite data” and the “infinite data”?

2)     Equation (10) should be revised. It differs from Equation (11). (z or z-1?)

3)     X-axis error in Figure 4(b).

Author Response

Please see the attachment.

Author Response File: Author Response.docx