Ultrafast laser bone ablation towards high surgical speeds and clinically relevant operation times for spine surgeries

Abstract

Recent advancements in ultrafast laser ablation technology redefine surgical precision while minimizing thermal damage, offering a promising alternative to traditional methods. However, the slow material removal rates (MRRs) have hindered clinical adoption. Addressing this challenge, we present a compact fiber-based laser delivery system, exhibiting an 82-fold increase in MRR compared to previous femtosecond laser probes. The system leverages a hollow-core Kagome fiber to deliver 10 ps laser pulses with high transmission efficiency and minimal nonlinear effects, even at high peak powers. The system distributes ultrashort pulses utilizing a piezo-scanned Lissajous-based beam steering mechanism onto the target surface over a larger field-of-view (FOV), enabling the scope for easy scalability of the system to a miniaturized probe. Drawing insights from our prior work, the focusing optics were carefully selected to deliver fluence three times the ablation threshold. An optimal combination of FOV size, translation speed, and repetition rate was identified, enabling clean ablations (devoid of carbonization) even at maximum laser power. We achieved a maximum MRR of 10.7 mm³/min with 8.8 W of laser power at 333 kHz, validating our hypothesis that high MRR can be achieved by ablating at high average powers over a large FOV with fast scanning of a large spot size. Numerical simulations further suggest that MRR up to 30 mm³/min can be achieved through increased repetition rates, expanded FOVs, and high translation speeds, defining an optimal parameter space for future probe designs. To validate the clinical relevance of high MRR, experiments were conducted to create deep bone incisions over a 3×3 mm² area within a clinically relevant timeframe. An ablation depth of ∼3 mm was achieved in ∼2 minutes without auxiliary cooling mechanisms. Scanning electron microscopy (SEM) of the deep incisions confirmed the preservation of healthy bone tissue, with clear evidence of canaliculi along the slopes and at the bottom surface of the ablated region. This study outlines a clear pathway toward developing a high-performance, miniaturized surgical probe with significant potential for spinal decompression surgery and other clinical applications, representing a transformative tool for future surgical precision.

1. Introduction

Ultrafast laser surgery has the potential to serve as an alternative minimally invasive surgical technique where precision is of utmost concern [1–6]. Bone surgery in the spine is one such sensitive surgical domain, where ultrafast lasers hold a very promising prospect [7,8]. The need for precise alternate strategies stems from the shortcomings of the current techniques in spine surgery and the promise of better patient care. The current techniques for generic bone surgeries involve mechanical tools such as ronguers, oscillating saws, and burs, which cause collateral damage to the surrounding tissues, leading to high complication rates and increased recovery times [8–10]. In specific cases such as the treatment of spinal stenosis, the nature of the medical procedure requires these tools to be inserted to reach the spinal column area multiple times before biting away a 1-3 mm length of bone spur [5]. Since such medical procedures are performed in a very tight and sensitive location of the spine, there are cases of collateral damage in the form of dural tears and nerve damage, which further hinder patient recovery [8,11–15]. Ultrafast laser ablation provides a higher degree of spatial and thermal confinement than the conventional lasers [16–19], and, thus, is an ideal surgical choice for high-precision spinal bone surgeries.

While ultrafast lasers are routinely used in ophthalmology [20–22], miniaturization of the laser delivery system and its associated optics is required for clinical adoption in other medical domains such as spinal surgery. Miniaturization requires fiber delivery of ultrafast laser pulses with high peak intensities, which invokes the use of specialized fibers and custom-made optical components [23]. While recent efforts in miniaturization of ultrafast laser surgery systems have shown great promise for clinical translation of ultrafast laser surgery [2,3,5,24–31], there remains an important challenge for achieving clinically viable surgery times.

It is well understood that ultrafast laser ablations suffer from low material removal rates (MRR), primarily owing to its precision and the associated mechanisms, such as nonlinear absorption confined to the focal plane, leading to localized material removal [16,18,19,32]. Most research efforts in understanding the efficacy of laser ablation in bones have been restricted to the use of free-space, bulk-optical setups [1–3,25,30,33,34]. Gemini et al. [2] ablated porcine femur with 350 fs pulses for three different wavelengths (1030 nm, 515 nm, and 343 nm) using different combinations of average power, repetition rate, and scanning speeds. They report achieving a maximum MRR of 0.66 mm³/s in the visible regime for a 250 kHz repetition rate, with 6.27 W of average power. Interestingly, they also reported that ablations the visible regime is more susceptible to carbonization, thanks to the strong absorption of hemoglobin. Zhang et al. [25] achieved a MRR of 0.99 mm³/s using a 200 kHz femtosecond laser with an average power of 19.7 W while drilling on sheep shank bones. However, both studies had used very high peak powers of 63 MW and 383 MW, respectively.

While these studies offer insights into femtosecond laser-assisted bone ablations on a benchtop, the clinical translation of this technology is often overlooked. For clinical implementation of this technology, miniaturization is necessary and, by itself, poses a unique set of additional challenges, such as flexible laser delivery, undesired nonlinear effects in bodily fluids above the bone surface, beam steering mechanisms, miniaturized focusing optics, and opto-mechanical probe design. Specifically, high average laser powers are always associated with high peak powers that can potentially induce detrimental non-linear effects, such as self-focusing or self-phase modulation, and therefore become the limiting factor. In spinal surgeries using fiber-based approaches, there is a risk of self-focusing occurring in either the fiber or the surrounding liquid (blood or irrigation saline).

Towards the development of ultrafast laser surgical probes, our group and others have made considerable efforts over the past decade [4,5,24,35–39]. Each probe had been developed with cognizance of a specific application, overcoming the challenging aspects of miniaturization in each stage. Most recently, our group developed a spinal surgery probe, with an intended application for the treatment of spinal stenosis [5]. A Kagome fiber delivered ultrashort pulses from a high repetition rate fiber laser to a custom-built miniaturized CaF₂ objective producing a spot size of ≈1.96 µm across a 90 × 90 µm² field of view (FOV). The probe was designed to suit the ergonomics of a hand-held spinal surgery probe, from the ruggedness of its housing for a better grip to a ∼2.6 mm diameter of tapered tip for better endoscopic visibility. The ablation performance of the probe was tested on excised bovine ribs and a maximum MRR of ∼0.13 mm³/min was obtained, mainly limited by the available maximum laser powers (< 0.24 W). The authors further conducted a numerical simulation of laser surgery parameters considering all limiting factors and concluded that higher average laser powers, balanced with larger spot size, scanning larger FOVs, and faster translation speeds, would lead to higher MRR.

For experimental validation of these numerical predictions, we studied MRR of excised bone samples using a high average power ultrafast laser in a benchtop optical setup. For clinical relevance of this study, we used Kagome fiber for delivering high peak powers and piezo beam scanning to enable eventual miniaturization. We selected the focusing optics taking cues of the important conclusions from the aforementioned numerical simulations. This paper demonstrates surgical speeds (MRR) above 10 mm³/min, which is a substantial increase (∼82×) from the previous results of ∼0.13 mm³/min. To evaluate the surgical relevance of MRR, we created deep bone trenches and assessed if clinically relevant surgical targets can be achieved. The benchtop setup can be readily transformed into a probe-based system by integrating off-the-shelf miniaturized focusing optics. Towards the development of a next-generation surgical probe for spinal stenosis, the current benchtop setup demonstrates great potential.

2. Experimental approach and methods

The primary consideration in designing the experimental setup was its potential for miniaturization into a surgical probe to facilitate clinical translation. To achieve this goal, we implemented a fiber-based delivery mechanism with beam-steering enabled by a piezo actuator, utilizing an in-line architecture similar to our previous probe designs [4,5,35,36]. However, for this study, we used off-the-shelf bulk optics to focus the laser pulses at the tissue end, allowing us to systematically investigate the parameter space for maximizing MRR. For clinical relevance, particularly for spinal applications, we used bovine bones as our experimental model. Both bovine and porcine bones are widely used to validate spinal procedures and laser ablation studies [3,30,34,40–42] due to their anatomical, density, and molecular similarities to human vertebrae. We selected bovine rib bones for their availability from a local vendor and targeted the cortical region, which is stiffer than cancellous bone [43], yielding conservative ablation rates applicable to spinal contexts. These factors collectively make bovine ribs a suitable model for our study’s objectives.

2.1. Experimental setup and design considerations for optical parameters

The benchtop optical setup, as shown in Fig. 1(a), incorporates a Kagome hollow-core photonic crystal fiber with ∼ 52 µm inner core diameter, 36 µm mode field diameter (MFD), and fiber NA ≈ 0.02 (Fig. 1(b)) [4,44]. The fiber delivers high-peak power laser pulses from a Yb-doped ultrafast fiber laser source (40 W, λ = 1035 nm, 300 fs – 10 ps, single pulse – 50 MHz, Monaco 1035-40-40, Coherent Inc.) to the bone sample. The maximum pulse energy that the laser can provide is 40 µJ for repetition rates up to 1 MHz. The laser source is guided to the coupling lens (L1 in Fig. 1(a)) using anti-reflection coated ultrafast mirrors and two pairs of attenuators made of a half-wave plate (HWP) and polarizing beam splitter (PBS), after which the maximum available pulse energy is ∼34 µJ. The p-polarized laser is then coupled into the fiber using a plano-convex lens (f = 100 mm), achieving an overall coupling and transmission efficiency of 81.07% across a fiber length of ∼6 m (Fig. 1(c)). Factoring in the fiber transmission, a maximum of ∼27 µJ of pulse energy is available at the tissue-end.

Fig. 1.

Optical layout of the benchtop system and Kagome fiber characterization. (a) Schematic of the bench-top setup used for studying bone ablation experiments. The green-circled area shows the cross-sectional schematic of the hypodermic tube-piezo-fiber assembly, where; ① represents the fiber, ② the inserts holding the fiber and piezo, ③ the piezo (EBL#3 Piezoceramic Tube), ④ the insert holding piezo and hypodermic tube, and ⑤ the hypodermic tube. HWP: Half-wave plate (WPH05M-1030, ThorLabs); M1, M2, M3: Ultrafast mirrors (UM10-45B, ThorLabs); PBS: Polarizing beam splitter (PBS25-1064-HP, ThorLabs); BB: Beam block (LB1, ThorLabs); L1: Coupling lens (LA1986-B, ThorLabs); L2, L3: Focusing lenses in a 4f-configuration (L1: LA1509-B, L2: LA1131-B); TS: Translation stage (NanoMax 302, ThorLabs). (b) Photograph of the benchtop system highlighting its major components. ① represents a bundle of wires extending from the hypodermic tube-piezo-fiber assembly. The orange wire corresponds to the bifurcation tubing, which protects the Kagome fiber, while the red, blue, and black wires correspond to the piezo—two quadrants and the ground, respectively. (c) Measured loss and calculated dispersion spectra of Kagome fiber. The inset shows en face view of the fiber. (d) Measured transmission curve for a 6 m long Kagome fiber at 10 ps pulse duration and 333 kHz repetition rate. The laser is p-polarized at the fiber input. (e) Pulse compression ratio at the fiber output for 300 fs (black), 5 ps (blue), and 10 ps (red) pulses as a function of input pulse energy. Data points represent the FWHM pulse widths measured with the autocorrelator. Error bars represent 95% confidence intervals of sech² fits to autocorrelator measurements. Color-coded solid lines serve as visual guidance for the trend.

The choice of Kagome fiber stems from our primary requirement for the flexible delivery of laser pulses with very high peak intensities in a clinical setting. The selected Kagome fiber, thanks to its large air core size, can support high peak power delivery with minimal pulse dispersion (GVD < – 437 fs²/m) and minimal loss (attenuation < 0.1 dB/m) at 1035 nm (Fig. 1(b)). Previous experimental and numerical investigation of the same fiber demonstrated minimal pulse distortion for pulse widths of 5 ps and 1.1 ps, with input pulse energies up to 10 µJ [4]. Here, we further studied the fiber’s performance at higher pulse energies, available up to ∼34 µJ at the fiber input.

Specifically, we measured pulse duration compression through the fiber for various input pulse energies and pulse durations at a repetition rate of 333 kHz (Fig. 1(d)). Significant pulse compression was observed for input pulse energies exceeding 10 µJ for 5 ps pulses. In contrast, only minimal compression occurred with 10 ps pulses at an input energy of 33 µJ, resulting in an output pulse duration of ∼ 9.3 ps. These results highlight the capability of Kagome fibers to handle high peak powers while minimizing nonlinear effects.

Further, since surgeries in a clinical setting may involve bodily fluids, we estimated a minimum pulse width of $\tau \sim 8$ ps was necessary to stay below the critical power of self-focusing in water ( $P_{cr}$ = 3 MW) using the relationship $\tau =0.88E_{pulse}/P_{cr}$ [45], where $E_{pulse}=27$ µJ for our case. Given that our laser source provides variable pulse widths up to 10 ps, we selected 10 ps as the operating pulse width for added safety.

We also confirmed that, at the maximum pulse energy, the irradiance at the fiber input (≈ 0.3 TW/cm²) is two orders of magnitude lower than the air ionization threshold (≈ 40 TW/cm²) [46]. This analysis confirms that an operating pulse width of 10 ps is optimal, enabling ablation at very high average powers while avoiding significant nonlinear effects.

For beam scanning, we adapted a piezo-enabled beam steering mechanism to scan across the desired FOV [4,5,35,36]. At the distal end, as encircled in green in Fig. 1(a), the fiber is attached to the four-quadrant piezo ceramic tube using a thin diamond-turned insert, which allows the fiber to function as a cantilever. The entire fiber-piezo assembly is then secured to the outer hypodermic tube using an additional insert [4]. A cross-sectional view of the fiber-piezo assembly is presented in Fig. 1(a). Applying a calibrated potential difference across the four quadrants of the piezo induces deflection of the fiber along two perpendicular axes, facilitating beam scanning across the desired FOV.

Further along, we implemented a 4f-lens system to focus the laser beam onto the tissue surface with a 2× demagnification, utilizing off-the-shelf 1″ plano-convex lenses. The selection of 2×-demagnification was informed by the fact that the most efficient ablation could be achieved with fluences ∼3 times the threshold [5]. Another consideration was the variation of the ablation threshold with pulse width. Since the ablation threshold scales as ∼τ^0.5 for pulse widths between 4 ps and 8 µs [32], and the measured ablation threshold of bovine ribs is 1.43 J/cm² at 1.5 ps [5], we estimated a threshold of 3.69 J/cm² for 10 ps pulses. Therefore, our optimal fluence range is ∼11.07 J/cm². Using this fluence alongside the maximum available pulse energy at the tissue end ( $E_{pulse,max}$ $\approx$ 27 µJ), we estimated the minimum spot radius ( $w_o$ ) to be 8.81 µm. Given the fiber mode field radius is 18 µm (MFD ≈ 36 µm), we determined that a demagnification of ∼2× would be ideal.

Together with the 4f-system, the piezo scan enabled rendering the desired FOV on the tissue. To translate the probe across a specific region of interest (ROI) on the bone, we secured the bone sample on a microscope slide and mounted onto a 3-axis motorized stage (MAX302, Thorlabs) capable of controlled movement. The motorized stage allowed for translation speeds of up to 4 mm/s along the tangential plane (perpendicular to the optical axis). A three-dimensional control at the tissue-end enabled us to demonstrate the system’s capability to create deep and healthy trenches on the bone surface. During experiments, the exposed cortical side of the mounted bone sample was positioned near the 4f-system’s focal plane and exposed to very low laser powers. The focal plane was identified by moving the sample until a strong second harmonic generation (SHG) signal could be observed on the bone surface.

2.2. Optical setup characterization methods

We characterized the beam profile across the focal plane by translating the imaging setup, comprising of a 10× objective (0.28 NA, M Plan Apo SL, Mitutoyo), a tube lens (200 mm focal length), and a beam profiler camera (SP928, Ophir Photonics). We used the same setup to characterize the fiber’s resonant frequencies and deflection properties and the resulting FOV for different applied voltages. To further characterize the FOV size, we also ablated borosilicate glass coverslips and imaged the ablated region. During ablation, the focal plane was found by translating the coverslip into the focal region, characterized by a faint white plasma on the surface of the coverslip. We incorporated a shutter (SH1 (ThorLabs), with a response time of 10 ms with SC10 controller) between the fiber tip at the distal end and the first lens of our 4f-focusing setup to modulate the laser exposure.

2.3. Design of bone ablation experiments

We designed the bone-ablation experiments to validate our numerical hypothesis that rapid distribution of laser pulses over a large FOV on the tissue surface using a large spot size enables delivery of high average powers to increase MRR. We performed all experiments at 10 ps pulse duration, with a laser repetition rate of 333 kHz, and at the maximum FOV size available with the current setup (550 × 550 µm²). We evaluated the MRR for varying laser fluences and two translation speeds (2 mm/s and 4 mm/s). These ablation experiments allowed us to determine the fluence threshold at a 10 ps pulse duration. In the second part of the study, we aimed to ablate deep trenches at a large ROI (3 × 3 mm²), representing a clinically relevant size for spine decompression surgeries [47]. Following ablation, we characterized the surface of these deep trenches to assess the health of the remaining bone and the quality of the ablated surface.

2.4. Preparation of bone samples

We purchased freshly cut bovine rib bones from the local vendor and prepared the samples using the following protocol: (1) Soft tissues attached to the bone surface were removed using a scalpel, leaving only the cortical surface exposed. (2) The samples were washed in a solution of 70% ethyl alcohol and, after drying, polished using sandpapers of varying grits to create a flat and smooth surface. Polishing the bone samples, particularly with coarser grits, could embed grit particles into the bone’s pores, potentially acting as localized heat sources during laser ablation. To mitigate this issue, we minimized polishing, used fine grit sandpapers, and selected regions with naturally flatter surface profiles when available. This approach reduced the risk of debris-related effects while maintaining sufficient surface uniformity to avoid axial defocusing during translation across the region of interest, ensuring reliable and repeatable data for our study. (3) The polished bone samples were then epoxied onto a microscope glass slide, ensuring the flat polished cortical surface was exposed. All experiments were performed within 24 hours of procuring the bone samples.

2.5. Bone ablation trench characterization methods

Our experiments involved 3-5 bovine rib bone samples per condition. Each experimental condition was defined by specific combinations of FOV (500 µm scan width), repetition rate (333 kHz), translation speed (2 or 4 mm/s), and average fluence. We employed statistical power calculations to establish the minimum sample size required for our experiments. Preliminary data revealed mean ablation depths of 177 µm (8 W) and 205 µm (8.8 W) at a translation speed of 2 mm/s, with a standard deviation (SD) of 17 µm. To detect a significant difference (H₀: µ ≤ 177 µm, H₁: µ > 177 µm) with 95% significance and 90% power, we calculated that a minimum of five samples was sufficient using a one-tailed, one-sample t-test. Similarly, at a translation speed of 4 mm/s, we determined that four samples were adequate to identify a difference (H₀: µ ≤ 75 µm (8 W), H₁: µ > 75 µm (8.8 W)) with an SD of 7 µm, maintaining the same significance and power levels. Since accurately determining the maximum material removal rate (MRR) was a key objective, statistical power calculations focused primarily on high-power conditions (8 W and 8.8 W) at both translation speeds. For lower power levels, we deemed three samples per condition sufficient to accommodate biological variability. These sample size estimates are based on performing a single ablation per sample. However, since we performed at least 3 ablations per sample (to ensure intra-sample consistency), our 3–5 samples per condition exceeds these minimum requirements.

We measured the profiles of ablated trenches using an optical profilometer (Keyence VK-X1100). The characterized regions were selected away from the start/end of the trenches to avoid the areas of high number of overlapping pulses caused by the acceleration/deceleration phases of the travel. Figure 2(a) shows a representative isometric view of a ∼1.2 mm segment of a longer ablated trench ( $\ge$ 4 mm) as captured by the profilometer. For each segment, we analyzed at least three 400 µm-long sections, as shown in the averaged cross-sectional profile (Fig. 2(b)), using the profilometer’s software to calculate average ablation depth (measured at the trench center), width, and cross-sectional area. We determined the MRR using:

$$\text{MRR = Cross - Sectional Area}\times \text{Translation Speed}.$$ (1)

Fig. 2.

Representative analysis of single ablation trench. (a) An example 3-D rendered isometric view of an ablation trench, as measured using the optical profilometer. The trench was created using laser pulses with an average power of 7 W, operating at 333 kHz repetition rate, scanned over a 550 $\times$ 550 µm² FOV at a 2 mm/s translation speed. (b) Cross-sectional profile of an averaged 400 µm section of the trench shown in (a). Key parameters, including average trench depth, width, and cross-sectional area were extracted from these profiles. (c) Scanning electron microscope (SEM) image of a segment of the ablated trench shown in (a). (d) Numerical simulation of the Lissajous scan for two different translational speeds (2 mm/s and 4 mm/s) and a fixed FOV scan width of 550 µm. The simulations show near complete coverage for both speeds. The color bar represents the number of overlapping pulses ( $N$ ) across the FOV.

Reported values (e.g., ablation depth, width, MRR) represent averages across all replicates—encompassing multiple sections (≥ 3) from multiple trenches (≥ 3) across multiple bone samples (≥ 3) per condition. Error bars indicate the standard error of the mean, reflecting the precision of these averages. This hierarchical replication approach ensured reliable results by addressing both biological and experimental variability. Our statistical framework prioritizes reproducibility across biological and experimental replicates.

To assess potential thermal effects of ultrafast laser bone ablation, we analyzed the surface properties of selected samples, primarily those with deep trenches, using a scanning electron microscope (FEI Quanta 650 ESEM). To prevent charging during imaging, we coated the samples with a very thin conductive layer of gold-palladium film using a DC sputter. Special care was taken during SEM imaging to avoid melting the exposed surfaces. Figure 2(c) presents an example SEM image of a trench, demonstrating no thermal effects.

Interestingly, the pattern formed on the bottom surface of the trench closely resembles the pattern predicted by our Lissajous simulations (Fig. 2(d)). The Lissajous scan trajectory is defined by [48]:

$$x(t)=A_x[sin(2\pi f_{x}t)];y(t)=A_y[cos(2\pi f_{y}t)],$$ (2)

where $A_{x,}{A}_y$ coefficients are the amplitude of the waveform that physically equates to half the FOV scan width and $f_x$ and $f_y$ are the scanning frequencies along x- and y- directions, respectively. The cantilevered fiber tip naturally exhibits different x- and y- resonant frequencies owing to slight eccentricity caused by gravity, thereby fulfilling the minimum requirement for Lissajous scanning [48]. For our simulations, we set the experimental values for $f_x$ = 805 Hz, $f_y$ = 850 Hz, predicting the scan trajectory along a single transverse direction.

For both translation speeds shown in Fig. 2(d), the simulations achieved near-complete coverage (≥ 99%), theoretically assuring ablation across the entire FOV of 550 $\times$ 550 µm². These simulations also provided an estimate of the average number of overlapping pulses along the mid-plane of the FOV, which was used to characterize ablation and numerically estimate ablation depth and MRR.

3. Results and discussion

3.1. Optical setup characterization

Figure 3(a) shows representative beam profiles at the focal plane, along with an inset displaying the beam quality as imaged on the beam profiler. The results indicate slight variations in the $1/e^2$ spot radius ( $w_o$ ) between the sagittal and longitudinal directions. This inhomogeneity is likely caused by slight pressure exerted on the fiber tip by its holder in the light coupling setup. For all fluence estimates, we used the averaged value, $w_o=$ 9.46 µm, which corresponds to an overall demagnification ratio of ∼1.9, given the mode-field radius of the fiber is ∼18 µm [4,44]. The average Rayleigh length was determined to be $z_R=$ 248.2 ± 29.9 µm.

Fig. 3.

Optical characterization of the benchtop system. (a) Focal spot after the 4f-system imaged using a 10× objective, 175 mm tube lens, and beam profiler. A representative beam profile has been given as an inset. Beam profile was fit to a Gaussian function to determine the $1/e^2$ focal spot radius ( $w_o$ ). Uncertainties represent 95% confidence interval of the fit along the longitudinal and sagittal axes of the beam. (b) Field of view (FOV) characterization, involving imaging with a 4× objective, 175 mm tube lens, and beam profiler (top images), as well as ablations of borosilicate glass coverslips (bottom images). Ablation FOVs agree within 5% of the target FOV size. Scale bars represent 200 µm.

Figure 3(b) presents the FOV characterization using borosilicate glass ablation, along with the beam profiler images. The beam profile images show a piezo-enabled Lissajous-scan pattern targeting FOVs of 400 $\times$ 400 µm² and 550 $\times$ 550 µm² by differentially actuating the fiber along the tangential and longitudinal axes at predetermined voltages and frequencies. Corresponding FOVs of 396 $\times$ 415 µm² and 541 $\times$ 530 µm², both within 5% of their respective target FOVs, were obtained in glass ablations.

3.2. Bone ablation performance at high average powers

Using the selected operating parameters of 333 kHz repetition rate, 10 ps pulse duration, and a 550 $\times$ 550 µm² FOV, we performed experiments to determine the MRRs for various combinations of translation speeds (2 mm/s and 4 mm/s) and average powers. We ablated trenches ( $\ge$ 4 mm) at different average powers, ranging between 4 W and 8.8 W, corresponding to average fluences of 4.3 J/cm² and 9.4 J/cm². Figure 4 illustrates the characteristics of the ablation trenches, including trench depth, trench width, and the corresponding MRR as a function of increasing fluence.

Fig. 4.

Evaluation of maximum achievable MRR on bovine cortical bone with 550 × 550 mm² FOV and 333 kHz laser repetition rate for two translation speeds. (a) Measured ablation depths, (b) trench widths, and (c) MRRs as a function of the average laser fluence. The maximum measured MRR is 10.7 ± 0.8 mm³/min for the translation speed of 2 mm/s at 8.8W average power (9.4 J/cm²). All the data points are an average of ≥ 3 measurements of ablation depths and widths. The solid lines in (a) represent the linear fits to data points based on Eq. (3). The solid lines in (c) represent the numerically estimated MRR based on the fluence thresholds and optical penetration depths determined from the linear fits in (a). Error bars represent the standard of the mean.

3.2.1. Ablation depth and threshold measurements

Figure 4(a) presents the ablation depth measurements, which exhibit a linear relationship with the logarithmic fluence, as expected. This trend is consistent across both translational speeds. For a constant FOV and repetition rate, a slower translational speed results in a higher pulse overlap, leading to greater ablation depth compared to faster translational speeds. Notably, Fig. 4(a) demonstrates that the ablation depth at 2 mm/s is nearly double that observed at 4 mm/s.

Additionally, we used the data from Fig. 4(a) to estimate the ablation threshold based on the linear relationship between the ablation depth per pulse ( $d$ ) and the logarithm of average fluence ( $F_{avg}$ ) (Fig. 4(a)), as described by:

$$d={\alpha }_{\mathit{eff}}^{-1}\ln \left(\frac{F_{avg}}{F_{th}}\right),$$ (3)

where ${\alpha }_{eff}^{-1}$ the effective optical penetration depth, as defined by Beer-Lambert’s law, and $F_{th}$ is the fluence threshold [49–51]. Considering the incubation effect that occurs when multiple laser pulses overlap, the ablation depth ( $h_a$ ) has the following relationship with ablation rate [50,52–54]:

$$h_a=d(N-N_{th})$$ (4)

Here, $N_{th}$ is the minimum number of pulses required to initiate ablation. For fluences greater than the ablation threshold, as in our case, $N_{th}$  = 0, leading to the following governing equation for $h_a$ :

$$h_a={\alpha }_{\mathit{eff}}^{-1}\ln \left(\frac{F_{avg}}{F_{th}}\right).$$ (5)

Here, N represents the number of overlapping pulses along the center plane of the trenches, which can be estimated using Lissajous simulations. The slope of the linear dependence between $h_a$ and $\ln (F_{avg})$ yields $N{\alpha }_{\mathit{eff}}^{-1}$ , while the slope-to-intercept ratio gives $F_{th}$ . The results of these calculations are summarized in Table 1. As expected, the measured fluence thresholds (3.17–3.58 J/cm²) are in close agreement with our theoretical estimate of 3.69 J/cm².

Table 1. Summary of bone ablation parameters extracted from linear fits of average ablation depth and the logarithm of average fluence for two translation speeds for data presented in Fig. 4(a). The table summarizes the estimated average number of overlapping laser pulses (N) along the trench mid-plane (determined from Lissajous scanning simulations), effective penetration depth ( ${\alpha }_{\mathit{eff}}^{-1}$ ), and ablation fluence threshold ( $F_{th}$ ). Uncertainties are based on the standard error of the mean of the slopes and y-intercepts.

$TranslationSpeed$ (mm/s)	$N$	${\alpha }_{\mathit{eff}}^{-1}$ (µm)	$F_{th}$ (J/cm2)
2	56	3.12 ± 0.26	3.17 ± 0. 53
4	28	3.10 ± 0. 28	3.58 ± 0.59

3.2.2. Material removal rate (MRR) measurements

Figure 4(b) shows the trench width, which increases slightly with the increasing fluence and plateaus at ∼575 µm beyond 6.5–7 J/cm², highlighting minor FOV curvature. Figure 4(c) provides the MRR values, which increase linearly with fluence, mainly a consequence of the corresponding increase in ablation depths (Fig. 4(a)). We achieved a maximum MRR of 10.7 ± 0.8 mm³/min at an average fluence of 9.4 J/cm² (∼8.8 W), which is equal to approximately 3 $F_{th}$ . This MRR value represents an ∼82× increase compared to our previous studies where we had a limited laser power (∼34× less), representing a ∼2.4× increase in the ablation efficiency, namely MRR per 1 W of average power. Such an increase in efficiency may be related to the measured increase in the optical penetration depth (4×) for the longer laser pulse durations used in this study.

This substantial increase in MRR confirms our hypothesis [5] that ablating over a large FOV with sufficiently high laser powers, combined with scanning using a large spot size, can significantly enhance MRR. A large spot size is necessary for maintaining fluences at an optimal value ( $\sim 3F_{th})$ , while large FOVs are crucial to limit the number of overlapping pulses below a specific threshold. Ablation is expected to saturate (cease) when the number of overlapping pulses exceeds a maximum limit, at which point the fluence drops below the $F_{th}$ value, as described below. Employing large FOVs and/or fast translation speeds effectively prevents this ablation saturation by modulating the distribution of pulses, which would otherwise reduce the efficiency of ablation process.

Additionally, we conducted a post-hoc analysis to determine the statistical significance of our findings. A one-way ANOVA on ablation depth and material removal rate (MRR) for both translation speeds revealed significant differences among the means for each power level (p < 0.0001). Tukey's HSD test, applied at α = 0.05, identified significant differences in ablation depths at 8 W and 8.8 W for 2 mm/s (adjusted p = 0.0057), MRR at 8 W and 8.8 W for 2 mm/s (adjusted p = 0.0119), ablation depths at 8 W and 8.8 W for 4 mm/s (adjusted p < 0.0001), and MRR at 8 W and 8.8 W for 4 mm/s (adjusted p < 0.0001). These results confirm that our sample sizes and experimental design were sufficient to detect significant differences, supporting the conclusions drawn from the data.

A further increase in MRR beyond 10 mm³/min can be achieved by increasing the average power further while keeping the peak power below $P_{cr}$ , which is feasible by increasing the repetition rate. At a repetition rate of 1 MHz, the MRR values are expected to triple by simply tripling the average powers. To understand if MRR can be further increased using our current laser parameters, we performed numerical estimations for a range of FOVs and translation speeds using the empirical parameters measured in this study.

Briefly, we estimated the number of overlapping pulses using the Lissajous scanning simulations described in Eq. (2) (Fig. 5(a)). We determined the theoretical ablation depth, $h_a$ , using Eq. (5), where N is estimated from Lissajous simulations and the ablation threshold and optical penetration depths at 10 ps are taken from Table 1 (Fig. 5(b)). The MRR was finally determined (Fig. 5(c)) using:

$$\mathrm{M}\mathrm{R}\mathrm{R}=\mathrm{F}\mathrm{a}\mathrm{c}\mathrm{t}\mathrm{o}\mathrm{r}\times (\text{FOV Scan Width})\times (\text{Translation Speed})\times h_a,$$ (6)

where the Factor (0.77) was estimated by fitting a linear regression to the available data for cross-sectional area versus the product of FOV scan width and $h_a$ .

Fig. 5.

Numerical estimates of MRR at 1 MHz repetition rate using 27 W of average power. (a) Number of overlapping pulses, (b) ablation depth, and (c) MRR for different combinations of FOV scan widths with a focal beam radius of 9.46 µm and 27 µJ of pulse energy (9.6 J/cm²) at tissue surface.

Figure 5 represents the numerical estimates for 1 MHz at 27 W average power, using the experimentally determined spot radius of 9.46 μm. The results show that MRR values approximately 30 mm³/min can theoretically be achieved for a wide range of FOV scan widths and translation speeds. Interestingly, MRR values remain constant ~30 mm³/min, over a large range of translation speeds (4–14 mm/s) and FOV scan widths (500–800 μm). This finding is particularly valuable in the context of the surgical tool, as it indicates that a consistent relative performance can be expected regardless of the operating parameters. Consequently, this consistency reduces the need for surgeons to optimize these variables during surgery, simplifying the operation process. The dotted black lines in Fig. 5 indicate the maximum achievable ablation depth, which is limited by a threshold value defined as

$$z_{th}=z_r\sqrt{{\left(\frac{w_{th}}{w_o}\right)}^2-1},$$ (7)

where $w_{th}$ is the $1/e^2$ beam radius along the optical axis for which $F_{avg}=F_{th}$ . It is evident that excessing the overlapping pulse threshold ( $N_{th}=113$ ) at high repetition rates results in reduced ablation efficiency. This result underscores the importance of employing large FOVs and/or high translation speeds to mitigate excessive pulse overlap. These findings lay the foundation for the design of our forthcoming probe and future experiments.

3.3. Ablating deep incisions in large areas

Our efforts to enhance MRR are focused on developing a clinically relevant ultrafast laser surgery system for treating spinal stenosis. Our vision is to create a handheld tool capable of swiftly creating deep incisions in bone spurs, facilitating their removal with a single tool insertion. Typically, bone spurs range from 1 to 3 mm in size. A single-level bone decompression during a lumbar laminectomy typically takes 20–40 minutes when using mechanical tools, setup time excluded [47]. Our goal is to minimize complications like soft tissue injuries or dural tears, which are often caused by the use of sharp cutting tools [11,55] while also ideally reducing operation times. Before proceeding with the design and development of a miniaturized surgical tool, we must first identify and test the parameter space necessary to create deep incisions in the bone. Through our experiments, we achieved the highest MRR using an 8.8 W average power while translating at a speed of 2 mm/s. Therefore, these parameters are our most favorable choice to achieve the desired clean ablation (devoid of carbonization) results within a clinically significant timeframe.

To create deep incisions, we ablated the bone samples by translating them along all three axes and repeating the ablation process over a select ROI of 3 $\times$ 3 mm². Scanning a FOV of 550 $\times$ 550 µm² for a length of 2.5 mm on the bone surface at 2 mm/s resulted in trenches ∼3 mm long and ∼550 µm wide. To cover the entire ROI, each longitudinal ablation was followed by a transverse translation of 500 µm. Once the entire ROI was scanned, an axial translation of 200 µm (equivalent to the ablation depth at the trench center) was performed to maintain the surface within the focal plane. Subsequently, the laser was backtracked, completing the process with an axial translation at the starting point. For ablating deep incisions, we pseudo-raster scanned the FOV over the entire ROI and implemented axial translations at the start and end points. To avoid delivering many thousands of overlapping pulses in a single spot during the slow axial translation, we used a mechanical shutter to block the laser exposure for 350 ms until the axial translation was completed. We inducted this latent period after each longitudinal translation. In summary, the shutter remained open for 1.25 s during each longitudinal translation (2.5 mm translation at 2 mm/s) and closed for 0.35 s during each transverse and axial translation to prevent excessive pulse deposition during slow axial movement. Therefore, each cycle, namely the total duration between each axial translation, took approximately 9.6 s. The whole process of translation in all three axes, including shutter control, was fully automated and repeated for a predetermined number of cycles to achieve deep bone incisions.

To characterize the deep incisions, we measured the average depth and width of the cross-sectional profiles of the ablated regions using the optical profilometer (Fig. 6(a)). The incision depths increased with numbers of cycles, while the bottom trench dimensions decreased in each cycle, resulting in a ∼60% reduction at the bottom of the 2.8-mm deep incision (Fig. 6(b)). Such reductions have previously been attributed to various phenomena, including diffraction of laser pulses at the edges, influencing the distribution of energy in subsequent pulses and altering ablation efficiency at the edges [50,56,57] and reflection or scattering of laser beam at the ablated wall surfaces due to Fresnel reflection [58].

Fig. 6.

Incisions as deep as 2.8 mm were achieved within 2 minutes over a large area of 3 × 3 mm² with operating parameters of 8.8 W, 333 kHz, 550 $\times$ 550 µm² FOV, and 2 mm/s. (a) A 3D rendered isometric view of a deep incision, as observed from the optical profilometer. (b) Averaged cross-sectional profiles of 400 µm sections of the deep incisions created with different number of scans. SEM image of (c) the top surface and (d) the bottom surface of the deepest incision made after 14 scans. Clear evidence of canaliculi along the slopes and on the bottom surface implies healthy ablations. Video of the ablation process for 2.8 mm incision can be found in Visualization 1 ^{(9.1MB, mov)} .

Incision depths of 2.1 mm and 2.4 mm only took less than 2 minutes of laser exposure. We could obtain an incision depth reaching 2.8 mm within ∼2 minutes and were devoid of any thermal effects as evident from the SEM images presented in Figs. 6(c) and 6(d). We used SEM imaging to evaluate the health of the incisions, specifically by examining the presence of bone canaliculi, which are essential for osteocyte viability [59,60]. These microscopic structures serve as key regulators of various bone functions [61], and disruptions in the canalicular network can lead to osteocyte apoptosis and impaired bone remodeling [60]. Preservation of the canalicular network post-ablation signifies minimal thermal impact, making it a strong indicator of “healthy ablations.” The SEM images show that these canals are intact along the incision edges, with no signs of damage from ablation, confirming the high quality of the cut. SEM investigations of these incisions only allude to the fact that ultrafast ablations indeed have the potential to ablate deep without the assistance of any cooling mechanism—a significant finding for miniaturization, as it eliminates the need for extra cooling components.

4. Conclusions

We successfully demonstrated that the MRR of bone samples could be significantly increased by delivering high average power ultrafast laser pulses at high repetition rates and balancing the distribution of overlapping pulses through scanning a large spot size over a wide FOV. These results were achieved using a fiber-based, piezo-scanned benchtop system, which holds the potential miniaturization and clinical adoption. Drawing on the insights gained from our previous study [5], we designed the focusing optics in a standard 4f-optical configuration to have a FOV of 550 × 550 µm² and a spot radius of 9.46 µm. A hollow-core Kagome fiber was used to deliver high energy pulses to the tissue-surface. Fiber characterization studies revealed negligible compression for 10 ps pulses, even at an input pulse energy of ∼34 µJ. This result underscores the potential of such fibers in delivering high peak powers while avoiding significant nonlinear effects, making them ideal for clinical applications.

The ablation potential of the benchtop system was characterized through parametric studies on excised bone tissue samples. Ablation at near maximum available laser powers of 8.8 W (9.4 J/cm² at 333 kHz) produced carbonization-free, clean cuts at translation speeds of 2 mm/s and 4 mm/s. We achieved an MRR of 10.7 mm³/min, representing 82× improvement compared to our previous study [5]. The measured ablation thresholds of bovine cortical bone, as well as the optical penetration depths, were higher than previously reported values for similar hard tissue ablations. This outcome was expected, since we selected to use 10 ps pulse widths, which offered a safe margin for ablating at high powers while staying below the critical power of self-focusing in water above the tissue surface [5]. One interesting observation was that the laser-induced incisions exhibited irregular or jagged walls, as shown in Figs. 2(a) and 2(b). We believe that these characteristic features may stem from the inherent heterogeneity of bone tissue, particularly its layered architecture in the cortical region [62], which likely causes a differential response to laser ablation, resulting in undulations and jagged profiles along the incision walls. However, the precise mechanism remains unclear, warranting future investigation.

To determine the maximum MRRs that can be achieved in ultrafast laser surgery, we performed additional numerical simulations considering factors such as fiber delivery, miniaturization, and the critical power limitations for self-focusing in water in the presence of bodily fluids ( $P_{peak}$  < 3 MW), and assuming $F=3F_{th}$ for efficient ablation. The simulations estimated that an MRR $\approx$ 30 mm³/min can be achieved by ablating at a 1 MHz repetition rate for various combinations of a wide range of FOV scan widths (500–800 µm) and translation speeds (4–14 mm/s). Further increase in ablation rates can be realized by delivering higher average powers using multiple fibers/spot sizes and/or by utilizing burst-mode ultrafast laser ablation methods [28]. Additionally, it may be noted that while ablations at 8.8 W (9.4 J/cm², 333 kHz) with translation speeds of 2 and 4 mm/s and FOV scan width of 500 µm yielded carbonization-free cuts, exceeding this power, for example to 12–14 W, without adjusting FOV, translation speed, and repetition rate may induce carbonization due to excessive thermal accumulation. Our prior experience indicates a power threshold, dependent on these parameters’ combination, beyond which carbonization occurs. This threshold varies with experimental conditions, and future studies will aim to explore these limits systematically across a broader parameter space.

To further demonstrate the clinical relevance of the achieved high MRR, we created deep bone incisions by scanning a large ROI on the bone surface and translating the sample along all 3 axes. We were able to achieve ablation depths of ∼3 mm within ∼2 minutes over a 3 × 3 mm² ROI. SEM images of the top and bottom surfaces of the ablated incisions revealed healthy bone tissue, highlighting the potential of ultrafast laser ablations to be performed without the need for cooling aids. With an appropriate miniaturized objective design and knowledge gained in the development of multiple miniaturized ultrafast laser surgery and imaging probes over the past 20 years [4,5,35–38,63], the benchtop system presented here can be adapted into a fiber-based, miniaturized surgical probe, having the potential of a clinically viable tool for spinal decompression surgeries.

Funding

National Institutes of Health 10.13039/100000002 ( NIH R01-DC014783); Fundação para a Ciência e a Tecnologia 10.13039/501100001871 ( UTA18-001217).

Disclosures

The authors declare no conflicts of interest.

Data availability

Data underlying the results presented in this paper are not publicly available at this time but may be obtained from the authors upon reasonable request.