Identifying Optimal da Vinci Tool Orientations for Photoacoustic Guided Hysterectomies€¦ ·  · 2017-09-21INTRODUCTION Approximately 600,000 hysterectomies ... tomies typically

Identifying Optimal da Vinci Tool Orientations forPhotoacoustic Guided Hysterectomies

Margaret Allard,aJoshua Shubert,b Muyinatu A. Lediju Bell* b,c

aSmith College, Department of Physics, Northampton, MA, USAbJohns Hopkins University, Department of Electrical and Computer Engineering, Baltimore,

MD, USAcJohns Hopkins University, Department of Biomedical Engineering, Baltimore, MD, USA

August 26, 2017

ABSTRACT

In hysterectomies, the uterus is removed and one of the possible complications is that the surgeon mayaccidentally damage the ureter because it is only a few millimeters away from the uterine arteries. We suggestusing photoacoustic imaging in da Vinci hysterectomies to detect the ureter with our new light delivery system,so that surgeons can more easily avoid it. Our light delivery system surrounds the da Vinci tool with opticalfibers. Images were taken by sweeping the tool across our uterine vessel model which was covered in ex vivotissue that was between the model and the fiber as well as between the ultrasound transducer and the model.To analyze the difference between images taken with different tool orientations, images and their correspondingda Vinci tool tracking coordinates were taken. Contrast and background SNR were also measured. The optimaltool orientation was determined based on how these measurements reflected the quality of the images taken. Theoptimal tool orientation had a contrast of over 10 dB from at least one centimeter away and a mean backgroundSNR of over 1.8, meaning there was minimal clutter in the image. It was also found that distance and angleaway from initial arm orientation greatly impacted what part of the vessel was being viewed, so when hoping toimprove an image, the angle and orientation should not be changed.

Keywords: minimally invasive surgery, robotic surgery, photoacoustic image guidance, surgical navigation,hysterectomy

1. INTRODUCTION

Approximately 600,000 hysterectomies (i.e., surgical removal of the uterus) are performed each year in theUnited states, and approximately 1 in 3 women over 60 will undergo this procedure in her lifetime.1 Hysterec-tomies typically follow the onset of medical conditions such as endometriosis (where cells that are supposed togrow inside the uterus grow outside of it), uterine prolapse (where the uterus starts collapsing into the vagina),and uterine cancer.

To remove the uterus, the surgeon must cut and cauterize the uterine arteries while avoiding the ureter, thetube from the kidneys to the bladder, which is located a few millimeters from the uterine arteries.2 If the ureteris injured and the damage is recognized intraoperatively, surgical time is increased, resulting in a heightened riskof infection. However, approximately 50-70% of uretal injuries are undetected during surgery,3 which results ina more severe prognosis. The longer the injury is undetected after the surgery, the more likely the developmentof kidney failure.

Hysterectomies are trending toward performance with robotic assistance, particularly with the da Vinciteleoperated surgical robot, due to increased dexterity, decreased hospital stays, 3-dimensional stereoscopicvisualization, minimal blood loss, and generally shorter recovery periods4–6 (although some studies indicate thatoperating times and length of hospital stay are not always improved7). While the da Vinci robot is used forother types of minimally invasive surgeries, such as radical prostatectomy,8 cardiac surgery,9,10 thyroid surgery,11

*E-mail:[email protected]

1

Figure 1: Proposed photoacoustic method for real-time imaging of the ureters and uterine arteries

and thoracic surgery,12 one of the most common uses of the da Vinci robot is performing minimally invasivehysterectomies.

Currently, the primary information available for uretal avoidance during robotic hysterectomies is the sur-geon’s knowledge of general patient anatomy, the surgeon’s experience, and the surface view provided by theendoscope of the da Vinci. This combination of information provides the surgeon with a general idea about thelocation of the ureter and the surrounding uterine artery, but some patients may have ureters that are moredifficult to detect than others.

This clinical challenge can potentially be addressed with the assistance of real-time image guidance. Whileultrasound imaging is a potential option, it would be difficult to constantly maneuver an ultrasound probe to findthe arteries and the ureter during surgery. In addition, the ureter and the uterine arteries are both hypoechoictargets that would have similar appearance in a real-time ultrasound image obtained with a drop-in probe.

We propose to use photoacoustic imaging to visualize the ureter and uterine artery during hysterectomies.Optical fibers that surround a da Vinci tool would illuminate the surgical site. The uterine arteries, whichhave higher optical absorption than surrounding tissue, would absorb this light, undergo thermal expansion, andgenerate a sound wave that could be detected with a transvaginal ultrasound probe. This concept is illustratedin Fig. 1. Because urine has a low optical absorption,13–15 our overall vision includes contrast agents for uretervisualization. If a biocompatible contrast agent that is only sensitive to a narrow band of wavelengths16 is insertedinto the urinary tract, the ureters can also be visualized with photoacoustic imaging, when the wavelength of thelaser is tuned to the optimal wavelength of the contrast agent. With this approach, the surgeon can potentiallyhave more information about the relative positions of the ureter and the uterine arteries. In addition, becausemetal has a high optical absorption coefficient, the da Vinci tool can also be visualized in the photoacousticimage if it is located within the image plane.

The primary contributions of this paper include demonstrating the feasibility of robotic photoacoustic-guidedhysterectomies by developing and testing a custom light delivery system to surround a da Vinci scissor tool andinvestigating the optimal wrist orientations of the scissor tool. These tasks were explored with a 3D printedmodel of the uterine artery surrounded by ex vivo bovine tissue to provide the optical and acoustic scatteringthat would be caused by surrounding tissue in a hysterectomy procedure.

2. METHODS AND MATERIALS

2.1 Experimental Setup

Our experiments were performed in a mock operating room that contained a da Vinci S robot, consisting ofan endoscope to visualize the surgical field, a master console (shown on the right of Fig. 2), and patient side

2

Figure 2: Photograph of the experimental setup

manipulators (shown on the left of Fig. 2), which are teleoperated from the master console. Only one of thepatient side manipulators was used for our experiments, although three of these robot arms are shown in Fig. 2.

Our photoacoustic imaging system was positioned next to the operating table which contained our exper-imental phantom (described in more detail below). The photoacoustic imaging system was comprised of anAlpinion ECUBE 12R ultrasound system attached to an Alpinion L3-8 linear transducer and a Phocus MobileLaser with a 1-to-7 fiber splitter17 attached to the 1064-nm output port. The 7 fibers of this light delivery systemsurrounded a da Vinci curved scissor tool and they were held in place with our custom designed, 3D printedfiber holder. The da Vinci scissor tool was held by one of the patient side manipulators of the da Vinci S robot.

Our custom modular phantom (used in previous work18) was built from laser cut acrylic pieces (held in placewith silicone glue) and 3D printed components. To simulate the uterine arteries, a 3D model of the arteriesaround the uterus was designed and 3D printed with black resin. This model was suspended by string throughthe holes of the phantom, and it is shown in Fig. 2, on the monitor displaying the endoscopic camera video feed.

The phantom was filled with water to permit acoustic wave propagation. The ultrasound transducer wasfixed against the acoustic window of the phantom and held by a Sawyer robot (Rethink Robotics) to act as astable passive arm for the experiments and to ensure that all images were acquired in the same image plane. A1.5 mm thick layer of ex vivo bovine tissue was draped over the phantom (as shown in the inset of Fig. 2), toreside in between the optical fiber and the vessels, and another layer of this same tissue was placed inside thephantom, between the 3D model and the transducer. These tissues were placed to introduce both optical andacoustic scattering for photoacoustic imaging.

2.2 Exploring Variations in Tool Orientations

The surgeon uses a wide range of surgical tools during the hysterectomy procedure. One required tool tosever the uterine artery is the curved scissor tool. This and other tools are used with dexterity that is similarto the human wrist motion. The wrist of the tool can therefore have multiple orientations during surgery, whichcould potentially impact the quality of photoacoustic images if a significant portion of the light is blocked by thetool.

We explored variations in photoacoustic imaging with the four wrist orientations of the curved scissor toolshown in Fig. 3(a). Orientation 1 shows no bending of the wrist with the scissors closed. Orientation 2 is the

3

(a)

(b)

Figure 3: (a) Photographs of tool orientations 1 through 4, from left to right, respectively. (b) Correspondinglight profiles for tool orientations 1 through 4, from left to right, respectively, acquired with 635 nm wavelengthlaser interfaced with the 1-to-7 fiber splitter.

same as Orientation 1, but the scissors are open. The wrist is bent in Orientation 3 and the joint connecting thescissors are also bent. In Orientation 4, the wrist is not bent, but the joint connecting the scissors is bent. Theseorientations permitted passage of a varied range of the light, as shown in the photographs Fig. 3(b), taken witha 635-nm laser light coupled to the input of the 1-to-7 fiber splitter. When performing the experiments with1064 nm wavelength pulsed laser light, the measured output energy per pulse for Orientations 1, 2, 3, and 4 was1.44, 1.44, 1.40, and 1.36 mJ, and the average input energy was the same for all orientations.

2.3 Tracking Tool Positions and Orientations

The da Vinci robot arm with the custom light delivery system attached to the curved scissor tool was sweptaway from the portion of the 3D model that was designed to mimic the portion of the uterine artery that crossesthe ureter. This sweeping motion was proposed and demonstrated in previous work.18,19 Because the phantomwas covered in tissue, we used the photoacoustic image appearance of the uterine artery to deterimine the startingpoint of each sweep, and we stopped sweeping when the vessel was no longer visible in the photoacoustic image.Once we found the signal in a photoacoustic image, we used our knowledge of the underlying hidden vessellocation to determine the direction of sweeping.

Photoacoustic images were acquired during each sweep for each tool orientation. The da Vinci robot kinemat-ics were utilized to track the position and orientation of the tool, and we simultaneously acquired this trackinginformation with each image acquisition. The position coordinates correspond to the position of the tool wristand the orientation corresponds to the orientation of the da Vinci robot arm. These tracking coordinates arepresented relative to the 3D vessel solid model in Fig. 4 where each trajectory is mapped in a different color andthe pink dots represent the start of each sweep. The sweep for Orientations 1, 2, 3, and 4 contain 16, 19, 6, and19 acquisition points, respectively. Orientation 3 had the fewest acquisition points because it was particularlydifficult for the user to maintain this orientation during the sweep. With the exception of Orientation 1, thedescribed sweeping motion was teleoperated from the master console of the da Vinci robot.

2.4 Data Analysis

To determine if the amount of light that the scissor tool blocks (relative to Orientation 1) affects our ability tosee portions of the uterine artery, the percentage of this vessel visible in each photoacoustic image was measuredand grouped by the associated tool orientation. These measurements were summarized with box plots, where thehorizontal line inside each box displays median error. The upper and lower edges of each box represent the firstand third quartiles of the data set. The vertical lines connected to the boxes show the minimum and maximum

4

Figure 4: Trajectories of the scissor tool wrist with axes representing the wrist position and the da Vinci armorientation.

Figure 5: Effect of d and θ on vessel visibility. (a) When the distance, d from a starting point changes and thereis no change in θ, the orignal field of view remains the same. (b) When there is a change in θ and a small changein d, there is some overlap in the field of view. (c) However, when there is a change in both d and θ, there is asignificant change in the field of view.

values in each data set, excluding outliers, which are shown as dots and defined as any value greater than 1.5times the interquartile range.

Based on our prior knowledge that distance from a target can diminish target contrast,19,20 we additionallymeasured the contrast of each signal as a function of distance from the starting point for each orientation.Contrast was measured as

Contrast =µvessel

µbackground(1)

where µvessel is the mean signal within a region of interest (ROI) inside the vessel signal (obtained by clickingalong the vessel with a computer mouse and expanding the pixel measurement to ±20 pixels (±0.4 mm) in theaxial dimension) and µbackground is the mean of the signals in the background ROI, defined to be the same sizeas the ROI, but translated by a distance of 0.6 mm to the right of the vessels. These measurements were madeusing the beamformed RF data and the ROIs were only defined once for each tool orientation.

As shown in Fig. 5, our ability to see a target (i.e., our optical field of view) is expected to be affected byboth distance and tool orientation. The light emitted from each of our seven optical fibers surrounding the toolhas a conical shape whose width is dependent on the numerical aperture of the fiber.17 Because of the conical

5

light profile geometry, if we do not change our tool orientation and we move away from the target by a Euclideandistance, d, we have a greater chance of seeing the same target (Fig. 5(a)). Similarly, we have greater toleranceto variations in changes in the angle that defines the orientation of our tool in 3D space, θ, if we are close to thetarget (Fig. 5(b)). However, the farther we move away from a target, the greater the impact variations in toolorientation will have on our ability to see the same target (or the same potion of a target, as illustrated in Fig.5(c)). Thus, we define a new metric, dθ, to determine if the same part of the vessel model was being imagedwith each sweep. As it relates to our experiments, d is the distance from the pink start point in Fig. 4 and θ isthe angular difference between the orientation axis at the first location on the pink dot.

Finally, based on our knowledge that the surgical tool tip will generate a photoacoustic signal and that thesesignals will manifest as clutter when the tool is not aligned with the ultrasound image plane, we measured thebackground signal-to-noise ration (SNR) of each acquired photoacoustic image and grouped these images by toolorientation. SNR was measured as

SNR =µbackground

σbackground, (2)

where σbackground is the standard deviation of the signals within the ROI defined as everything below theappearance of the vessel (where clutter artifacts from the tool are most likely to appear) and µbackground wascomputed for this same ROI. These measurements were performed on the beamformed RF data.

3. RESULTS

3.1 Vessel Visibility

To summarize vessel visibility for each tool orientation, the length of a vessel as it appeared in each imagewas measured and normalized by the greatest length measurement overall, which corresponded to the top leftimage in Fig. 6(a) (acquired with the scissor tool in Orientation 1). We therefore considered this vessel to bevisualized at 100%. Fig. 6(a) shows examples of images obtained with the remaining three tool orientations (asindicated above each figure). The boxplot in Fig. 6(b) shows the distribution of vessel visibility percentages foreach tool orientation. Orientation 1 visualized the most of the vessel, which is intuitive because it blocks theleast light. Orientation 3 had the lowest median of the four orientations, and it blocks the most light.

(a) (b)

Figure 6: Images taken with tool Orientations 1-4 (indicated above each image). (b) Boxplot showing the percentof the vessel visible in each orientation

6

It initially appeared as if Orientation 1 achieved 100% vessel visibility because it permitted the passage ofmore light. While it may be generalized that the vessel visibility is related to the percentage of light that isblocked by the tool, this generalization is not consistent across all results. For example, Orientation 4 blockedmore light than Orientation 2, but it produced images with greater vessel visibility, which indicates that thereare additional factors to consider when characterizing vessel visibility.

3.2 Contrast and Distance

For each tool orientation, the measured contrast was plotted as a function of distance from the start of eachsweep, as shown in Fig. 7(a). The four orientations were generally capable of producing high contrast images(which is considered to be within the range of 10-20 dB). When images were acquired at a distance greater than1 cm from the starting position, the contrast tended to drop below 5 dB, which is considered low contrast. Thecontrast measurements for each orientation were fit to third order polynomials to demonstrate this trend, withOrientation 3 presenting an exception to these findings.

In order to understand what caused Orientation 3 to have a high contrast at 2 cm and a low contrast at 0cm, both distance and angle were considered. Larger distances from the starting point are expected to be moreaffected by small angular changes, thus distance was plotted against our newly defined variable dθ in Fig. 7(b).We manually confirmed that images acquired with dθ > 0.2 were never looking at the same part of the vesselas the image acquired at the beginning of each sweep. For the first low contrast images of Orientation 3, theimages were visualizing the longer line of the vessel, and in the higher contrast images around 2 cm distance, thebifurcation point of the vessel was visualized. This result indicates that even though the early images were closeto the start, they were far from the long vessel branch. Similarly, the images acquired with Orientation 3 fartheraway from the start were viewing a portion of the vessel that was closer to the tool (based on Fig. 4, where theend trajectory of Orientation 3 is closer to the vessel bifurcation point). This result also confirms that dθ is animportant metric to consider when characterizing vessel visibility with each sweep (assuming that the same partof the vessel is initially visualized).

3.3 Image Clutter

When light from the optical fiber is absorbed by the metal of tool and the tool tip is outside of the imageplane, the presence of acoustic clutter from the tool tip could complicate image interpretation. As each toolorientation absorbs varying degrees of light (based on the light profile images in Fig. 3(b)), tool orientation couldimpact the amount of image clutter present in an image, particularly if the tool tip is outside of the image plane.Fig. 8 (a) is an example of an image with minimal clutter acquired with the tool in Orientation 1, while Fig. 8

(a) (b)

Figure 7: (a) Contrast measurements were plotted as a function of distance and fit using third order polynomials.(b) The distance measurements were plotted as a function of our newly defined dθ term and a dotted was addedto show the separation point between images that were looking at the same region of the 3D model as that ofthe initial starting point for each image. This demarcation was visually confirmed for each photoacoustic image.

7

(b) shows an image with more clutter from the tool tip, acquired with the tool in Orientation 4. We know thatthe extra signals are caused by the tool tip because the tool tip was outside of the image plane (as confirmed byultrasound images), the signals only appeared when a significant portion of tool tip was in the path of the laser,and the real-time photoacoustic images showed that the location of these extra signals moved when the tool tipmoved. We extended the image depth to fully capture and characterize the impact of these extra signals thatappear deeper in the image.

Background SNR was measured to quantify clutter in the image and plotted in Fig. 8(c). Images acquiredwith Orientation 1 had the least clutter and highest background SNR with a mean of 1.8. Orientation 2 producedimages with slightly more clutter and a mean background SNR of 1.7, while Orientations 3 and 4 produced imageswith more clutter and mean background SNRs below 1.5.

(a) (b) (c)

Figure 8: Photoacoustic images acquired with (a) Orientation 1, containing minimal acoustic clutter and (b)Orientation 2, containing acoustic clutter from the out-of-plane tool tip. (c) Bar graph summarizing the mean± one standard deviation of background SNR measured in all images from each tool orientation

4. DISCUSSION

To determine if there is an optimal tool orientation for the surgeon to use for photoacoustic imaging, vesselvisibility, contrast, and clutter were investigated. At first, it seemed likely that orientation determined thepercentage of the vessel that could be viewed in a photoacoustic image. However, it was clear that each imagewas not only viewing different lengths of vessel, but different sections of the vessel itself. Once the trajectoriesof the sweeps for each orientation were compared to the images, it became clear that image sweeps that weretaken on the left, farther down the line of the vessel, visualized a larger portion of the vessel. Based on theseresults, it was decided that the percentage of the vessel visible is more tied to the location of the tool and not thepercentage of light being blocked. So, it was found that vessel visibility was not determined by tool orientation,so that did not go into the final consideration for which orientation is optimal.

Contrast was measured to see if different tool orientations could produce images where the signal could beeasily detected. It was found that all orientations were capable of producing high contrast images above 10dB when at least 1 cm away from the vessel. It was also determined that small changes in theta at non-zerodistances can result in a great change in the part of the vessel that is being visualized. This means that if thesurgeon finds a signal at any orientation and wants to improve it, they should lock the theta, and then approachthe target.

Image clutter needs to be minimal because clutter from out of plane tools could be mistaken for the toolitself. This means that a surgeon could look at a cluttered image and erroneously think they know the locationof the tool. It was found in the results that orientation 1 produced images with the least clutter. Orientation 2was slightly worse, and orientations 3 and 4 produced images with significant clutter.

Due to all of these factors, it was decided that orientation 1 is the optimal tool orientation for creatingphotoacoustic images for surgical guidance. This is because orientation 1 was found to produce high contrast

8

images when at least 1 cm away from the beginning of the sweep, and it produced the least clutter. Allorientations can produce helpful photoacoustic images, but the images produced with orientation 1 are the leastconfusing due to their lack of significant clutter. In fact, if a signal is found while in a different orientation (notorientation 1), to improve the found signal, the surgeon is advised to lock theta before approaching in order tomaintain the part of the vessel visualized in the image.

5. CONCLUSION

We demonstrated the feasibility of integrating photoacoustic imaging with the da Vinci robot in order toavoid the ureter and improve targeting of the uterine arteries during hysterectomies. Our integration included aspecialized light delivery system to surround a da Vinci curved scissor tool. We additionally provide a detailedanalysis of the optimal tool orientations for photoacoustic-guided surgeries using a scissor tool that partiallyblocks the transmitted light, indicating that the four orientations investigated have the potential to producesufficient images for photoacoustic guidance. The optimal orientation involved no bending of the tool’s wrist,thus, if a surgeon desires a clear photoacoustic image of the uterine artery or ureter with minimal confusionabout the tool location, the best option is to straighten the tool’s wrist (Orientation 1). However, to avoid losingsight of a low-contrast signal, it is helpful to lock all angular degrees of freedom before approaching this signalto improve its contrast (instead of adjusting the wrist to achieve the optimal orientation).

Disclosures

The authors have no relevant financial interests in this manuscript and no potential conflicts of interest todisclose.

Acknowledgements

This work was completed in partnership with the NSF Computational Sensing and Medical Robotics ResearchExperience for Undergraduates program. Funding was provided by NSF Grant EEC-1460674 and NIH GrantR00-EB018994. The authors thank Anton Deguet for assistance with obtaining da Vinci tracking coordinatesand Formlabs Inc. (Somerville, MA) for their 3D printing assistance.

REFERENCES

1. National Center for Health Statistics (US). Advance data from vital and health statistics. Number 16-20. USDept. of Health and Human Services, Public Health Service, Centers for Disease Control and Prevention,National Center for Health Statistics, 1995.

2. Lawrence H Bannister and Peter Llewellyn Williams. Gray’s anatomy: the anatomical basis of medicineand surgery. Churchill Livingstone, 1995.

3. Scott E Delacroix and JC Winters. Urinary tract injures: recognition and management. Clinics in colonand rectal surgery, 23(02):104–112, 2010.

4. Arnold P Advincula. Surgical techniques: robot-assisted laparoscopic hysterectomy with the da vinci R©surgical system. The international Journal of medical robotics and computer assisted surgery, 2(4):305–311,2006.

5. S Maeso, M Reza, JA Mayol, JA Blasco, M Guerra, E Andradas, and MN Plana. Efficacy of the davinci surgical system in abdominal surgery compared with that of laparoscopy: a systematic review andmeta-analysis. Annals of surgery, 252(2):254–262, 2010.

6. Todd M Beste, Keith H Nelson, and James A Daucher. Total laparoscopic hysterectomy utilizing a roboticsurgical system. JSLS: Journal of the Society of Laparoendoscopic Surgeons, 9(1):13, 2005.

7. Enrique Soto, Yungtai Lo, Kathryn Friedman, Carlos Soto, Farr Nezhat, Linus Chuang, and Herbert Gretz.Total laparoscopic hysterectomy versus da vinci robotic hysterectomy: is using the robot beneficial? Journalof gynecologic oncology, 22(4):253–259, 2011.

8. Ashutosh Tewari, James Peabody, Richard Sarle, Guruswami Balakrishnan, Ashok Hemal, Alok Shrivastava,and Mani Menon. Technique of da vinci robot-assisted anatomic radical prostatectomy. Urology, 60(4):569–572, 2002.

9

9. W Randolph Chitwood, L Wiley Nifong, Joseph E Elbeery, William H Chapman, Robert Albrecht, VictorKim, and James A Young. Robotic mitral valve repair: trapezoidal resection and prosthetic annuloplastywith the da vinci surgical system. The Journal of Thoracic and Cardiovascular Surgery, 120(6):1171–1172,2000.

10. L Wiley Nifong, Victor F Chu, B Marcus Bailey, David M Maziarz, Vincent L Sorrell, Donald Holbert, andW Randolph Chitwood. Robotic mitral valve repair: experience with the da vinci system. The Annals ofthoracic surgery, 75(2):438–443, 2003.

11. Sang-Wook Kang, Seung Chul Lee, So Hee Lee, Kang Young Lee, Jong Ju Jeong, Yong Sang Lee, Kee-Hyun Nam, Hang Seok Chang, Woong Youn Chung, and Cheong Soo Park. Robotic thyroid surgery usinga gasless, transaxillary approach and the da vinci s system: the operative outcomes of 338 consecutivepatients. Surgery, 146(6):1048–1055, 2009.

12. J Bodner, H Wykypiel, G Wetscher, and T Schmid. First experiences with the da vinci operating robot inthoracic surgery. European Journal of Cardio-thoracic surgery, 25(5):844–851, 2004.

13. Shangyuan Feng, Weiwei Chen, Yongzeng Li, Guannan Chen, Zufang Huang, Xiaohua Liao, Zhiming Xie,and Rong Chen. Surface-enhanced raman spectroscopy of urine by an ingenious near-infrared raman spec-trometer. In Photonics Asia 2007, pages 682628–682628. International Society for Optics and Photonics,2007.

14. Meng-Cai Huang, Hong-Wei Sun, et al. Study of normal and cancerous urine using photoacoustic spec-troscopy. Journal of biomedical engineering, 12(5):425–428, 1990.

15. SG Guminetsky, Olga V Pishak, Vasyl P Pishak, and PM Grigorishin. Absorbing and diffusive propertiesof blood plasma and urine proteins. In International Conference on Correlation Optics, volume 3317, pages390–398, 1997.

16. Akram Abuteen, Saeid Zanganeh, Joshua Akhigbe, Lalith P Samankumara, Andres Aguirre, NrusinghBiswal, Marcel Braune, Anke Vollertsen, Beate Roder, Christian Bruckner, et al. The evaluation of nir-absorbing porphyrin derivatives as contrast agents in photoacoustic imaging. Physical Chemistry ChemicalPhysics, 15(42):18502–18509, 2013.

17. Blackberrie Eddins and Muyinatu A Lediju Bell. Design of a multifiber light delivery system forphotoacoustic-guided surgery. Journal of Biomedical Optics, 22(4), 2017.

18. N. Gandhi, M. Allard, S. Kim, P. Kazanzides, and M.A. Lediju Bell. Photoacoustic-based approach tosurgical guidance performed with and without a da vinci robot. Journal of Biomedical Optics, 22(12), 2017(accepted).

19. Sungmin Kim, Youri Tan, Peter Kazanzides, and Muyinatu A Lediju Bell. Feasibility of photoacousticimage guidance for telerobotic endonasal transsphenoidal surgery. In IEEE International Conference onBiomedical Robotics and Biomechatronics, 2016.

20. Muyinatu A Lediju Bell, Anastasia K Ostrowski, Ke Li, Peter Kazanzides, and Emad M Boctor. Localizationof transcranial targets for photoacoustic-guided endonasal surgeries. Photoacoustics, 3(2):78–87, 2015.

Appendices

1) My lab supported this topic of ethics by always examining our assumptions and making sure we wereproperly representing our data. Also, expectations about authorship were always addressed early.

2) In this REU, for the first time I was given my own research project where I was responsible for conductingexperiments to create publishable results. Also, I was able to help out on another project, resulting in me gettingpublished for the first time. I now have a much greater understanding of how to do data analysis and my Matlabcoding has greatly improved. I also was given the opportunity to write this own paper, so I learned how tostructure an academic paper as well.

3) If I were to recommend this program to a friend, I would tell them that it was a wonderful experiencewhere I was closely mentored, worked hard, and learned a lot, both about my lab’s area of study and about whatbeing a graduate student is like. I’m very grateful that I was able to be in this program and do work that couldreally help people with mentors and other students that wanted me to succeed.

10