Pocket MUSE: an affordable, versatile and high-performance fluorescence microscope using a smartphone

Smartphone microscopes can be useful tools for a broad range of imaging applications. This manuscript demonstrates the first practical implementation of Microscopy with Ultraviolet Surface Excitation (MUSE) in a compact smartphone microscope called Pocket MUSE, resulting in a remarkably effective design. Fabricated with parts from consumer electronics that are readily available at low cost, the small optical module attaches directly over the rear lens in a smartphone. It enables high-quality multichannel fluorescence microscopy with submicron resolution over a 10× equivalent field of view. In addition to the novel optical configuration, Pocket MUSE is compatible with a series of simple, portable, and user-friendly sample preparation strategies that can be directly implemented for various microscopy applications for point-of-care diagnostics, at-home health monitoring, plant biology, STEM education, environmental studies, etc.

smartphone microscope) are shown. The quality of the pseudo H&E images are not close to standard H&E slide images at all. With the data showing only what pocket MUSE can do, no comparison statement can be made. In IHC section (Fig.3), what actually was done is a pocket MUSE imaging of thick brain slice with antibody staining. This is not IHC in commonsense (paraffin slides with specific staining protocol), extremely confusing terminology usage. Furthermore, the processing is radically different from conventional IHC processing, and the appearance is also completely different. It is not possible to compare the images to typical IHC even for those who are familiar with IHC slides. Furthermore authors failed to include enough details of the special staining process in manuscript ("slightly higher concentration with overnight staining" Not specific enough to reproduce the experiment.) The authors also claim "Compared to conventional bright-field imaging, Pocket MUSE reveals more of .... (lines 176-177)", but what's actually shown is the comparison of pocket MUSE with and without UV excitation. The images shown as the bright-field images in the manuscript are not close to the quality of conventional bright-field smartphone microscope images because of the difference of tissue preparation and likely different optimization of optics. The statement is clearly wrong, and the result doesn't demonstrate the image quality relative to any of the standard/modalities. Figure.7, too, does show the comparison of pocket MUSE with and without UV. Their statement "pocket MUSE is better than conventional bright-field imaging" is not demonstrated, and the data doesn't serve as an evidence of their high quality imaging.

Baseless statements in the Discussion
The supplementary data is insufficient and not convincing to support the statement in line 223-226. Comparison of costs between preliminary smartphone microscopes and pocket MUSE has to be shown to support the statement at lines 231-232. The statement (fine alignment is not needed for MUSE because of the difference between MUSE and TIR) is not supported in the supplement note (254-255). No comparison with pocket MUSE vs histology are shown (line 259-260), and psuedo H&E images are far from good as standard H&E. In 266, 'the excellent IHC performance' was not shown in the paper as explained above.

Lack of conciseness
The manuscript has too much insignificant information, and fails to concisely bring key concepts. For example, in materials and methods, there are many insignificant details described such as the details of sanding. On the other hand, there are several key details such as specs of LEDs (power, driving voltage, current, and spectrum) are not shown or can be found only in supplementary materials. Also description of potential future applications of MUSE at the end of discussion section seems quite lengthy, given that feasibility of those are only minimally supported in the manuscript. (Fig 6-8) Also, there are too much materials in supplemental. The data used to justify key concept in the main text has to be moved to main text. Also information included to supplementary must be wisely selected and concisely edited. Authors must wisely highlight the key information rather than including every subtle details. It seems to me that many of the supplemental information should be submitted as a separate paper in an engineering journal.(as a personal suggestion)

Revise for readers of Communications Biology
The manuscript appeared to be a paper originally drafted for histology/medical journal, but is submitted to this biology journal with minor modification. There are too many of expressions "looking like" authors missed to change from the original draft rather than meant to be (like too often mentioned medical/histology as examples). Please read through again carefully and improve consistency. The intense optical/electrical engineering parts also will be needed to be simplified for general readers of Communications Biology.

incomplete/inappropriate citations
The choices of references are not optimized. About half of the references are for introducing 'hypothetical' application of the technology, not directly relate to the contents of manuscript at this point or even not imaging. In addition, references of key concepts such as MUSE, RACL and TIR are quite thin relative to those. Authors should review the choices of references and reorganize them. 6. Professional writing to public Authors should focus on showing results, describing methodology, discussing the findings and hypothesis concluded from evidences, rather than describing baseless comments or author-specific stories. For example, I think the first paragraph of supplemental 1.1 contains some inappropriate sentences to be shown in scientific articles.
In short, the technology is impressive and will be of interest to the readers, but there are a lot of works for authors to do to make the manuscript appropriate as a decent scientific writing. All reviewers found merit in our work but had some concerns before publication. We appreciate the suggestions to improve our paper. We have incorporated all the suggestions into the revised manuscript. Responses are highlighted in red. Changed text is highlighted in red in the manuscript.

Reviewer #1
Critique 1a: In lines 76-80, the authors mentioned using frustrated total internal reflection (TIR) as the light delivery. From the authors' drawing, it is still difficult to understand how the LED light goes. Response 1a: Lines 76-80 are in the introduction, where we did not feel detailed information was warranted. However, we further clarified the position of the LED in our results section (new lines 148-149) and thoroughly discussed "how the LED light goes" (e.g., positioning against the edge of the optical window) in the supplemental material (Supplemental Note 1.6).
To address this concern, we also added some brief introduction and referenced the supplemental material in the original lines 76-80 / new lines 81-83.
Critique 1b: Also, with the divergence of the LED light, what is the portion of light energy that can reach the sample through Frustrated TIR? Response 1b: We appreciate the important question about the portion of light reaching the sample. We didn't know a practical method to accurately measure the percentage of light reaching the sample in this setup. However, we did provide an in-depth discussion about LED coupling with geometric raytracing calculations in the supplementary materials (Supplemental Note 2.1-2.3). This discussion should effectively address the reviewer's question.
Briefly, the LED die is tightly held against the optical window's edge. Therefore, although the beam divergence is considerable, the spot size of the beam is just slightly larger than the size of the LED die at the edge of the optical window. A significant amount of light can be coupled into the optical window. The percentage amount of the light coupled into the optical window is dependent on the size of the LED die, the distance between the die and the edge of the optical window, and the angle of the edge of the optical window.
Critique 2: I thought the sample holder should be a quartz optical window for UV light delivery. However, in lines 148-150, the authors are using the word "glass" instead. Please double check. Response 2: Fused quartz is a common optical glass material, but the term "quartz" alone can be ambiguous since it is also a mineral. To clarify this, we changed the term "quartz" to "fused quartz" in new lines 97, 149, 437 & 438.
Critique 3: In general, there are not too many things to question. The reason is that the authors only show a collection of images. However, the usefulness of the images has not been shown, which is the weakness of the work. Therefore, I believe that the authors should, at the least, show one of the killing applications of Pocket MUSE to strengthen the significance of the manuscript. Response 3: Pocket MUSE is a device for taking microscopy images. To demonstrate the usefulness of Pocket MUSE images for histology, we added additional data to compare and characterize histology images taken with Pocket MUSE (e.g., new Figure 2, new lines 163-171, 579-585). We have also added references demonstrating the great potential of MUSE systems for preliminary histology imaging and biology education (new lines 284-286, 288). In the discussion of the paper, we also clarified the point that Pocket MUSE could extend these applications to locations where a fluorescence microscope is not available (new lines 286).

Critique 4:
In line 63, the authors mentioned that "to contribute to blur and background". I believe that it should be "to contribute to blur the background".

Reviewer #2
Critique 1a: The most novel aspect of the work is the FTIR illumination scheme. However, its effect on MUSE sectioning (axial resolution and contrast) isn't really discussed. From the original MUSE patents and the following experimental work of Yoshitake et al, the effect of angle of illumination is known to be critical to MUSE image resolution. While the manuscript provides detailed calculations and even ray tracing simulations of the FTIR process, the effect of angle of illumination and what effect it has on the pocket MUSE resolution is less clearly addressed. Response 1a: We thank the reviewer for bringing up this point. Because the angle of illumination affects the axial resolution and contrast of MUSE imaging, we actually added a full section in the supplemental material (Supplemental Note 2.7) covering how frustrated TIR affects axial resolution. Critique 1c: It should be simple to calculate the effect on axial resolution of the FTIR illumination and compare to conventional illumination (particularly in light of Yoshitake et al), which would greatly aid in evaluating the advantages of FTIR illumination. However, from some simple calculations (which may be inaccurate), I suspect that the design optimized for coupling efficiency rather than reducing angle of refraction into the sample (frustrated angle of refraction into the sample ranges from ~14 to 90 deg, with most of the refracted angles being near ~70-90 deg). This suggests a tradeoff in resolution/contrast, and possible scope for further optimization. Response 1c: Unlike in the Yoshitake et al. paper, our optical setup is quite different (e.g., involves a much broader distribution of angles incident upon the sample). We are not aware of a method to accurately calculate/measure the exact axial resolution and optical sectioning of the frustrated TIR illumination because of the broad distribution of angles and sample-specific penetration depth. Although we do not report a predicted optical section thickness with our setup, we did add Supplemental Note 2.7 to discuss the effects of optical sectioning with frustrated TIR illumination based on the concepts presented in Yoshitake's paper.
The reviewer is correct that the design is optimized for coupling efficiency because the photon budget is limited in this application. We would like to note that the reviewer may have misunderstood that no light can reach the sample at any air-immersion angles (impossible through TIR) based on the angle ranges given in the critique. Considering the broader range of angles delivered by frustrated TIR, the overall effect is between air immersion and water immersion from the paper by Yoshitake et al. This discussion is covered in the new supplemental section, and referenced in new line 270-273. We believe the tradeoff in resolution/contrast is possible scope for further optimization in the future.
Critique 2a: Sample mounting only relies on surface tension to the quartz plate. Would this not result in stability issues, especially when some of your exposure times range in seconds? Response 2a: In most cases, exposure is <=1/3 s. For the iPhones, that is the upper limit of the shutter speed. For some samples, 1/15 s or faster shutter speed is sufficient at a relatively low camera gain (e.g., ~ISO 400 for the iPhone 6s). Within less than 250 ms exposure, stability was rarely a problem.
Some cameras in Android phones can expose longer, but that is only necessary when the signal is weak. As the reviewer suggested, we do experience some blurriness when the exposure time ranges in seconds. Avoiding high-frequency movement by supporting the smartphone against the chest or a wall is sufficient to eliminate the stability issues when long exposure times are used. It is also always easy to take a few images and select the one with the least blur. Some of this discussion is in the method/Pocket MUSE imaging section (new line 457-477).
Critique 2b: Would this not result in more surface irregularities (from solid tissue samples, despite best efforts in cutting flat surfaces with basic tools) which might result in the following? Response 2b: As the reviewer asked, we do see some image artifacts resulting from surface irregularities when the sample is large and rigid (e.g., plant root). However, it is relatively easy to cut a flat surface on these samples with razors. In contrast, soft samples are difficult to cut (e.g., unfixed tissue), but surface tension can flatten the surface effectively. Some difficulty comes from fixed animal tissues. Some mechanical support may be helpful for these types of tissues. This discussion is included in the method section/Pocket MUSE imaging section (new line 460-465).