Dual-modality optical diagnosis for precise in vivo identification of tumors in neurosurgery

Abstract

In neurosurgery, the precise diagnosis and treatment of tumor diseases are challenging to realize. Current clinical diagnoses lack fast and accurate intraoperative information. Therefore, the development of new methods and techniques to assist neurosurgeons intraoperatively is necessary. Optical diagnosis is a promising method to provide accurate information about biological tissues in a short time. Therefore, in this study, we proposed a dual-modality optical diagnostic method through point-to-face registration fusion in the optical system. We incorporated quantitative autofluorescence spectroscopy and optical coherence tomography (OCT) and evaluated our methods in an animal model. Methods: A mouse model consisting of 16 nude mice was built by injecting the mouse brains with human glioma cells. Preoperative bioluminescence imaging was used to evaluate the growth states of tumors and locate the tumor sites. Quantitative autofluorescence spectroscopy, which provided local biochemical information with single-point detection, and OCT, which provided relatively global structural information with en face mapping scanning, were combined using the point-to-face registration fusion method to provide precise diagnostic information for identifying the brain tumors. Postoperative pathology was performed to evaluate the sensitivity and specificity of optical diagnosis. Results: Ex vivo quantitative autofluorescence spectroscopy and OCT imaging were first performed in eight mice to acquire the optimal measuring parameters for tumor staging and identification. We then performed in vivo quantitative autofluorescence spectroscopy and OCT imaging. The results showed that tumor staging could be realized through quantitative autofluorescence spectroscopy, and fusion images could be used to precisely identify tumors. The autofluorescence spectral map, OCT en face map, and fused diagnostic map had average sensitivities of 91.7%, 86.1%, and 95.9% and specificities of 93.2%, 96.0%, and 88.7%, respectively, for tumor identification. Conclusion: The dual-modality optical point-to-face registration fusion method and system we proposed could provide both biochemical information and structural information. The in vivo experimental results validated that the sensitivity (95.9%) of the fused map was higher than that of either single diagnostic modality (86.1% or 91.7%). Tumor staging was realized through quantitative autofluorescence spectroscopy. The proposed method will be applicable to future intelligent theranostic systems and improve many clinical neurosurgeries.

Keywords: brain tumor; dual-modality optical diagnosis; neurosurgery; optical coherence tomography; quantitative autofluorescence spectroscopy.

Figure 1

The research flowchart of the proposed point-to-face registration-based dual-modality optical diagnostic method for precise identification of tumors in neurosurgery.

Figure 2

The configuration of the in vivo quantitative autofluorescence spectroscopy module.

Figure 3

Schematic of the SS-OCT module. A 2-D galvo scanner was used. BPD: balanced photodetector.

Figure 4

(A) The preprocessing procedure. (B) The en face map of an OCT image cube; the tumor position is shown in the 2-D mode.

Figure 5

The framework of quantitative autofluorescence spectroscopy and OCT information fusion based on the point-to-face registration fusion method.

Figure 6

(A) The diagram of cutting pathological sections. (B) The cross-sectional pathological section and its corresponding rectangle. (C) The pathological map.

Figure 7

The dual-modality optical diagnostic system containing the quantitative autofluorescence spectroscopy module and the OCT module.

Figure 8

Bioluminescence imaging after two weeks of tumor cell injection.

Figure 9

(A) EEM of the brain tissue. (B) EEM of the tumorous tissue. (C) Spectral curves at 380 nm excitation. The emitted photon numbers of normal tissues and tumorous tissues at the emission wavelength of 462 nm were approximately 4000 and 2000, respectively; the emitted photon numbers of normal tissues and tumorous tissues at the emission wavelength of 614 nm were approximately 2000 and 4000, respectively. (D) Spectral curves at 450 nm excitation. The emitted photon numbers of normal tissues and tumorous tissues at the emission wavelength of 550 nm were approximately 1500 and 800, respectively; the emitted photon numbers of normal tissues and tumorous tissues at the emission wavelength of 612 nm were approximately 1000 and 2000, respectively. (E) Distribution of R₁. (F) Distribution of R₂.

Figure 10

The relationship between the tumor stage and quantitative autofluorescence spectroscopy. (A-D) The average spectral curves and the corresponding pathology three, six, nine and 12 days after tumor cell injection. (E, F) The trends of the average spectral parameters R₁ and R₂ with increasing tumor stage.

Figure 11

(A) The intensity curves of the A-scan data. (B) A cross-sectional OCT image. (C) The intensity curves for the tissue surface and their exponential fitting curves. The blue curves in (A) and (C) correspond to the blue line in (B), and the red curves in (A) and (C) correspond to the red line in (B). The attenuation coefficients of the red curve and blue curve were 0.0012 and 0.0023, respectively. (D) One B-scan OCT image of the cube. (E) Flattened OCT data with a thickness of 50 pixels. The fixed-pattern noises are marked by white arrows.

Figure 12

The en face map of attenuation coefficients and other reference information. (A) Tissue upper surface. (B) The en face map of the OCT image cube. (C) The OCT image corresponding to the yellow line in (A). (D) Pathology corresponding to the image in (C). The tumorous regions are marked by arrows. The tumor margins on the tissue surface were visible to the naked eye and were marked by the surgeon using India ink, as shown by the red dashes in (A) and (B), and short green lines in (C) and (D).

Figure 13

Binary map processing based on different thresholds. (A) The target area in the en face map. (B) Binary maps under different thresholds. (C) Diagnostic sensitivity (blue)/specificity (red) data. (D) The ROC curve. The binarization was conducted on the region in the inset outlined by blue lines in (A). The largest area of the rectangle under the ROC curve appeared at the point where the threshold equaled 6.6.

Figure 14

The fused map of OCT attenuation information and autofluorescence spectral indexes. (A). In vivo autofluorescence spectral scanning. (B). The naked-eye view of the mouse brain. The target scanning area is marked by the red box. (C). In vivo OCT imaging. (D). The spectral maps of R₁ and R₂ values of the target areas. (E). The OCT en face map of attenuation. The area outlined by the blue box was further cropped and binarized as the final map. (F). The fused map of the OCT en face map and the spectral maps, overlapped by the real scanning area. The tumorous area identified by OCT is in yellow, the tumorous area identified by autofluorescence is in red, the intersection tumorous area of the two modalities is in orange and is outlined by the blue dashes, and the union area is indicated by the black outline.

Figure 15

The comparison between fluorescence imaging and our proposed method in the C6 rat model four and eight days after tumor cell injection. (A, B) The diagnostic results by fluorescence imaging. The tumorous regions are marked by the green dashes. (C, D) The fused maps overlapped by the white-light image. The tumorous regions are marked by the orange dashes.