Stratified assessment of an FDA-cleared deep learning algorithm for automated detection and contouring of metastatic brain tumors in stereotactic radiosurgery.

PURPOSE: Artificial intelligence-based tools can be leveraged to improve detection and segmentation of brain metastases for stereotactic radiosurgery (SRS). VBrain by Vysioneer Inc. is a deep learning algorithm with recent FDA clearance to assist in brain tumor contouring. We aimed to assess the performance of this tool by various demographic and clinical characteristics among patients with brain metastases treated with SRS. MATERIALS AND METHODS: We randomly selected 100 patients with brain metastases who underwent initial SRS on the CyberKnife from 2017 to 2020 at a single institution. Cases with resection cavities were excluded from the analysis. Computed tomography (CT) and axial T1-weighted post-contrast magnetic resonance (MR) image data were extracted for each patient and uploaded to VBrain. A brain metastasis was considered "detected" when the VBrain- "predicted" contours overlapped with the corresponding physician contours ("ground-truth" contours). We evaluated performance of VBrain against ground-truth contours using the following metrics: lesion-wise Dice similarity coefficient (DSC), lesion-wise average Hausdorff distance (AVD), false positive count (FP), and lesion-wise sensitivity (%). Kruskal-Wallis tests were performed to assess the relationships between patient characteristics including sex, race, primary histology, age, and size and number of brain metastases, and performance metrics such as DSC, AVD, FP, and sensitivity. RESULTS: We analyzed 100 patients with 435 intact brain metastases treated with SRS. Our cohort consisted of patients with a median number of 2 brain metastases (range: 1 to 52), median age of 69 (range: 19 to 91), and 50% male and 50% female patients. The primary site breakdown was 56% lung, 10% melanoma, 9% breast, 8% gynecological, 5% renal, 4% gastrointestinal, 2% sarcoma, and 6% other, while the race breakdown was 60% White, 18% Asian, 3% Black/African American, 2% Native Hawaiian or other Pacific Islander, and 17% other/unknown/not reported. The median tumor size was 0.112 c.c. (range: 0.010-26.475 c.c.). We found mean lesion-wise DSC to be 0.723, mean lesion-wise AVD to be 7.34% of lesion size (0.704 mm), mean FP count to be 0.72 tumors per case, and lesion-wise sensitivity to be 89.30% for all lesions. Moreover, mean sensitivity was found to be 99.07%, 97.59%, and 96.23% for lesions with diameter equal to and greater than 10 mm, 7.5 mm, and 5 mm, respectively. No other significant differences in performance metrics were observed across demographic or clinical characteristic groups. CONCLUSION: In this study, a commercial deep learning algorithm showed promising results in segmenting brain metastases, with 96.23% sensitivity for metastases with diameters of 5 mm or higher. As the software is an assistive AI, future work of VBrain integration into the clinical workflow can provide further clinical and research insights.

Local control of brain metastases with osimertinib alone in patients with EGFR-mutant non-small cell lung cancer.

PURPOSE: Although osimertinib has excellent intracranial activity in metastatic non-small cell lung cancer (NSCLC) with exon 19 deletion or L858R EGFR alterations, measures of local control of brain metastases are less well-reported. We describe lesion-level outcomes of brain metastases treated with osimertinib alone. METHODS: We retrospectively reviewed patients with EGFR-mutant NSCLC with untreated brain metastasis measuring ≥ 5 mm at the time of initiating osimertinib. Cumulative incidence of local recurrence in brain (LRiB) was calculated with death as a competing risk, and univariable and multivariable analyses were conducted to identify factors associated with LRiB. RESULTS: We included 284 brain metastases from 37 patients. Median follow-up was 20.1 months. On initial MRI after starting osimertinib, patient-level response was complete response (CR) in 11 (15%), partial response (PR) in 33 (45%), stable disease (SD) in 18 (25%) and progressive disease (PD) in 11 (15%). The 1-year cumulative incidence of LRiB was 14% (95% CI 9.9-17.9) and was significantly different in patients with a CR (0%), PR (4%), and SD (11%; p = 0.02). Uncontrolled primary tumor (adjusted hazard ratio [aHR] 3.78, 95% CI 1.87-7.66; p < 0.001), increasing number of prior systemic therapies (aHR 2.12, 95% CI 1.49-3.04; p < 0.001), and higher ECOG score (aHR 7.8, 95% CI 1.99-31.81; p = 0.003) were associated with LRiB. CONCLUSIONS: Although 1-year cumulative incidence of LRiB is < 4% with a CR or PR, 1-year cumulative incidence of LRiB is over 10% for patients with less than a PR to osimertinib on initial MRI. These patients should be followed closely for need for additional treatment such as stereotactic radiosurgery.

Randomized multi-reader evaluation of automated detection and segmentation of brain tumors in stereotactic radiosurgery with deep neural networks.

BACKGROUND: Stereotactic radiosurgery (SRS), a validated treatment for brain tumors, requires accurate tumor contouring. This manual segmentation process is time-consuming and prone to substantial inter-practitioner variability. Artificial intelligence (AI) with deep neural networks have increasingly been proposed for use in lesion detection and segmentation but have seldom been validated in a clinical setting. METHODS: We conducted a randomized, cross-modal, multi-reader, multispecialty, multi-case study to evaluate the impact of AI assistance on brain tumor SRS. A state-of-the-art auto-contouring algorithm built on multi-modality imaging and ensemble neural networks was integrated into the clinical workflow. Nine medical professionals contoured the same case series in two reader modes (assisted or unassisted) with a memory washout period of 6 weeks between each section. The case series consisted of 10 algorithm-unseen cases, including five cases of brain metastases, three of meningiomas, and two of acoustic neuromas. Among the nine readers, three experienced experts determined the ground truths of tumor contours. RESULTS: With the AI assistance, the inter-reader agreement significantly increased (Dice similarity coefficient [DSC] from 0.86 to 0.90, P < 0.001). Algorithm-assisted physicians demonstrated a higher sensitivity for lesion detection than unassisted physicians (91.3% vs 82.6%, P = .030). AI assistance improved contouring accuracy, with an average increase in DSC of 0.028, especially for physicians with less SRS experience (average DSC from 0.847 to 0.865, P = .002). In addition, AI assistance improved efficiency with a median of 30.8% time-saving. Less-experienced clinicians gained prominent improvement on contouring accuracy but less benefit in reduction of working hours. By contrast, SRS specialists had a relatively minor advantage in DSC, but greater time-saving with the aid of AI. CONCLUSIONS: Deep learning neural networks can be optimally utilized to improve accuracy and efficiency for the clinical workflow in brain tumor SRS.

Nonlinear Imaging using Object-Dependent Illumination.

Nonlinear imaging systems can surpass the limits of linear optics, but nearly all rely on physical media and atomic/molecular response to work. These materials are constrained by their physical properties, such as frequency selectivity, environmental sensitivity, time behavior, and fixed nonlinear response. Here, we show that electro-optic spatial light modulators (SLMs) can take the place of traditional nonlinear media, provided that there is a feedback between the shape of the object and the pattern on the modulator. This feedback creates a designer illumination that generalizes the field of adaptive optics to include object-dependent patterns. Unlike physical media, the SLM response can provide a wide range of mathematical functions, operate over broad bandwidths at high speeds, and work equally well at high power and single-photon levels. We demonstrate the method experimentally for both coherent and incoherent light.

Enhanced phase retrieval using nonlinear dynamics.

Historically, phase retrieval algorithms have relied on linear propagation between two different amplitude (intensity) measurements. While generally successful, these algorithms have many issues, including susceptibility to noise, local minima, and indeterminate initial and final conditions. Here, we show that nonlinear propagation overcomes these issues, as intensity-induced changes to the index of refraction create additional constraints on the phase. More specifically, phase-matching conditions (conservation of wave energy and momentum) induce an object-dependent resonance between the measured amplitudes and the unknown phase. The result is a non-classical convergence profile in the reconstruction algorithm that contains a zero crossing, where the observable minimum in amplitude error and the unobservable minimum in phase error align at the same iteration number. We demonstrate this convergence experimentally in a photorefractive crystal, showing that there is a clear rule for stopping iterations. We find that the optimum phase retrieval occurs for a nonlinear strength that gives minimal correlation between the linear and nonlinear output amplitudes, i.e. a condition that maximizes the information diversity between linear and nonlinear propagation. The corresponding algorithm greatly improves the conventional Gerchberg-Saxton result and holds much potential for enhancing other methods of diffractive imaging.

Efficient Structure Resonance Energy Transfer from Microwaves to Confined Acoustic Vibrations in Viruses.

Virus is known to resonate in the confined-acoustic dipolar mode with microwave of the same frequency. However this effect was not considered in previous virus-microwave interaction studies and microwave-based virus epidemic prevention. Here we show that this structure-resonant energy transfer effect from microwaves to virus can be efficient enough so that airborne virus was inactivated with reasonable microwave power density safe for the open public. We demonstrate this effect by measuring the residual viral infectivity of influenza A virus after illuminating microwaves with different frequencies and powers. We also established a theoretical model to estimate the microwaves power threshold for virus inactivation and good agreement with experiments was obtained. Such structure-resonant energy transfer induced inactivation is mainly through physically fracturing the virus structure, which was confirmed by real-time reverse transcription polymerase chain reaction. These results provide a pathway toward establishing a new epidemic prevention strategy in open public for airborne virus.

Terahertz polarization-sensitive rectangular pipe waveguides.

We propose square and rectangular pipe waveguides for low-loss THz waveguiding and polarization control. Different from common circular-symmetric THz fibers and waveguides, the proposed rectangular pipe waveguides successfully remove the transmission degeneracy of two orthogonal polarizations and possess polarization sensitivity to the guided THz waves. By measuring the attenuation spectra, we find that the polarization sensitivity depends on the structure of the pipe waveguides. With butt coupling method, it is easy to combine circular pipe waveguides and the rectangular ones.

High-sensitivity in vivo THz transmission imaging of early human breast cancer in a subcutaneous xenograft mouse model.

We performed in vivo THz transmission imaging study on a subcutaneous xenograft mouse model for early human breast cancer detection. With a THz-fiber-scanning transmission imaging system, we continuously monitored the growth of human breast cancer in mice. Our in vivo study not only indicates that THz transmission imaging can distinguish cancer from the surrounding fatty tissue, but also with a high sensitivity. Our in vivo study on the subcutaneous xenograft mouse model will encourage broad and further investigations for future early cancer screening by using THz imaging system.

Performance of THz fiber-scanning near-field microscopy to diagnose breast tumors.

Based on tissues from 20 female patients (mean age: 53 years; rang: 36-72 years), we examine the performance of a room-temperature-operated terahertz (THz) fiber-scanning near-field microscopy to diagnose slices of breast tissues. The specimens were frozen sliced and then measured in a thawed state without dehydration. We performed the imaging at 320 GHz. Our study indicates that images acquired in the THz transmission-illumination mode can all clearly distinguish breast tumor tissues from normal tissues without H&E staining. Due to its capability to perform quantitative analysis and to allow follow-up staining and traditional pathohistological analysis, our study indicates great potential of the THz fiber-scanning near-field microscopy for future automation, which is critical for fast and complete pre-screening on breast tumor pathological examinations and for assisting quick definition of the tumor margins during the surgical procedure such as breast-conserving surgery.

Terahertz pipe-waveguide-based directional couplers.

We experimentally demonstrate a terahertz (THz) leaky mode directional coupler for future THz applications. The proposed directional coupler comprises two square pipe waveguides. The coupling efficiency is investigated for different frequencies, polarizations, and core sizes. Rectangular pipe-waveguide-based directional couplers and the issue of insertion loss are also discussed. It is found that the THz directional coupler works most efficiently in the minimal-attenuation wavelength regime.

Bending loss of terahertz pipe waveguides.

We present an experimental study on the bending loss of terahertz (THz) pipe waveguide. Bending loss of pipe waveguides is investigated for various frequencies, polarizations, core diameters, cladding thicknesses, and cladding materials. Our results indicate that the pipe waveguides with lower guiding loss suffer lower bending loss due to stronger mode confinement. The unexpected low bending loss in the investigated simple leaky waveguide structure promises variety of flexible applications.