We present initial evidence of the SOLUS potential for the multimodal non-invasive diagnosis of breast cancer by describing the correlation between optical and standard radiological data and analyzing a case study.
A machine learning classification algorithm is applied to the SOLUS database to discriminate benign and malignant breast lesions, based on absorption and composition properties retrieved through diffuse optical tomography. The Mann-Whitney test indicates oxy-hemoglobin (p-value = 0.0007) and lipids (0.0387) as the most significant constituents for lesion classification, but work is in progress for further analysis. Together with sensitivity (91%), specificity (75%) and the Area Under the ROC Curve (0.83), special metrics for imbalanced datasets (27% of malignant lesions) are applied to the machine learning outcome: balanced accuracy (83%) and Matthews Correlation Coefficient (0.65). The initial results underline the promising informative content of optical data.
Significance: Diffuse optical tomography is an ill-posed problem. Combination with ultrasound can improve the results of diffuse optical tomography applied to the diagnosis of breast cancer and allow for classification of lesions.
Aim: To provide a simulation pipeline for the assessment of reconstruction and classification methods for diffuse optical tomography with concurrent ultrasound information.
Approach: A set of breast digital phantoms with benign and malignant lesions was simulated building on the software VICTRE. Acoustic and optical properties were assigned to the phantoms for the generation of B-mode images and optical data. A reconstruction algorithm based on a two-region nonlinear fitting and incorporating the ultrasound information was tested. Machine learning classification methods were applied to the reconstructed values to discriminate lesions into benign and malignant after reconstruction.
Results: The approach allowed us to generate realistic US and optical data and to test a two-region reconstruction method for a large number of realistic simulations. When information is extracted from ultrasound images, at least 75% of lesions are correctly classified. With ideal two-region separation, the accuracy is higher than 80%.
Conclusions: A pipeline for the generation of realistic ultrasound and diffuse optics data was implemented. Machine learning methods applied to a optical reconstruction with a nonlinear optical model and morphological information permit to discriminate malignant lesions from benign ones.
A multimodal instrument for breast imaging was developed, combining ultrasound (morphology), shear wave elastography (stiffness), and time domain multiwavelength diffuse optical tomography (blood, water, lipid, collagen) to improve the non-invasive diagnosis of breast cancer.
To improve non-invasively the specificity in the diagnosis of breast cancer after a positive screening mammography or doubt/suspicious ultrasound examination, the SOLUS project developed a multimodal imaging system that combines: Bmode ultrasound (US) scans (to assess morphology), Color Doppler (to visualize vascularization), shear-wave elastography (to measure stiffness), and time domain multi-wavelength diffuse optical tomography (to estimate tissue composition in terms of oxy- and deoxy-hemoglobin, lipid, water, and collagen concentrations). The multimodal probe arranges 8 innovative photonic modules (optodes) around the US transducer, providing capability for optical tomographic reconstruction. For more accurate estimate of lesion composition, US-assessed morphological priors can be used to guide the optical reconstructions. Each optode comprises: i) 8 picosecond pulsed laser diodes with different wavelengths, covering a wide spectral range (635-1064 nm) for good probing of the different tissue constituents; ii) a large-area (variable, up to 8.6 mm2 ) fast-gated digital Silicon Photomultiplier; iii) the acquisition electronics to record the distribution of time-of-flight of the re-emitted photons. The optode is the basic element of the optical part of the system, but is also a stand-alone, ultra-compact (about 4 cm3 ) device for time domain multi-wavelength diffuse optics, with potential application in various fields.
Time Domain Diffuse Optical Tomography (TD-DOT) at different wavelengths can be used to retrieve tissue reconstructing the components of a two-region system starting from self-normalized time-dependent measurements performed in reflectivity geometry over multiple wavelengths. The proposed method performs a fit of a limited number of tissues parameters providing a good quantification of the components’ concentrations by applying a FEM-based Diffusion approximation of the TD-DOT direct model.
Time Domain Diffuse Optical Tomography (TD-DOT) performed at multiple wavelengths can be used to non-invasively probe tissue composition. Then, tissue composition can be related to breast tissue and lesion type. Thus, TD-DOT could be used for therapy monitoring for breast cancer. We developed a software tool for multi-wavelength TD-DOT and performed a validation on meat phantoms to mimic tissue heterogeneity. An inclusion of different meat was exploited to mimic the presence of a lesion in the breast. Results show good localization of the inclusion, but poor quantification of the reconstructed breast composition. The use of a morphological prior constraint, providing information on inclusion geometry and position, significantly improves both localization and composition estimate.
Time domain Diffuse Optical Tomography (TD-DOT) non-invasively probes the optical proprieties of biological tissue. These can be related to changes in tissue composition, thus making TD-DOT potentially valuable for cancer imaging. In particular, an application of interest is therapy monitoring for breast cancer. Thus, we developed a software tool for multiwavelength TD-DOT in reflectance geometry. While the use of multiple wavelengths probes the main components of the breast, the chosen geometry offers the advantage of linking the photon flight time to the investigated depth. We validated the tool on silicon phantoms embedding an absorbing inclusion to simulate a malignant lesion in breast tissue. Also, we exploited the a priori information on position and geometry of the inclusion by using a morphological prior constraint. The results show a good localization of the depth of inclusion but a reduced quantification. When the morphological constraint is used, though, the localization improves dramatically, also reducing surface artifacts and improving quantification as well. Still, there is room for improvement in the quantification of the “lesion” properties.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.