Deep Learning Classification of Breast Cancer Tissue from Terahertz Imaging Through Wavelet Synchro-Squeezed Transformation and Transfer Learning.

Terahertz imaging and spectroscopy is an exciting technology that has the potential to provide insights in medical imaging. Prior research has leveraged statistical inference to classify tissue regions from terahertz images. To date, these approaches have shown that the segmentation problem is challenging for images of fresh tissue and for tumors that have invaded muscular regions. Artificial intelligence, particularly machine learning and deep learning, has been shown to improve performance in some medical imaging challenges. This paper builds on that literature by modifying a set of deep learning approaches to the challenge of classifying tissue regions of images captured by terahertz imaging and spectroscopy of freshly excised murine xenograft tissue. Our approach is to preprocess the images through a wavelet synchronous-squeezed transformation (WSST) to convert time-sequential terahertz data of each THz pixel to a spectrogram. Spectrograms are used as input tensors to a deep convolution neural network for pixel-wise classification. Based on the classification result of each pixel, a cancer tissue segmentation map is achieved. In experimentation, we adopt leave-one-sample-out cross-validation strategy, and evaluate our chosen networks and results using multiple metrics such as accuracy, precision, intersection, and size. The results from this experimentation demonstrate improvement in classification accuracy compared to statistical methods, an improvement to segmentation between muscle and cancerous regions in xenograft tumors, and identify areas to improve the imaging and classification methodology.

Hyperspectral terahertz imaging and optical clearance for cancer classification in breast tumor surgical specimen.

Purpose: We investigate the enhancement in terahertz (THz) images of freshly excised breast tumors upon treatment with an optical clearance agent. The hyperspectral imaging and spectral classifications are used to quantitatively demonstrate the image enhancement. Glycerol solution with 60% concentration is applied to excised breast tumor specimens for various time durations to investigate the effectiveness on image enhancement. Approach: THz reflection spectroscopy is utilized to obtain the absorption coefficient and the index of refraction of untreated and glycerol-treated tissues at each frequency up to 3 THz. Two classifiers, spectral angular mapping (SAM) based on several kernels and Euclidean minimum distance (EMD) are implemented to evaluate the effectiveness of the treatment. The testing raw data is obtained from five breast cancer specimens: two untreated specimens and three specimens treated with glycerol solution for 20, 40, or 60 min. All tumors used in the testing data have healthy tissues adjacent to cancerous ones consistent with the challenge faced in lumpectomy surgeries. Results: The glycerol-treated tissues showed a decrease in the absorption coefficients compared with untreated tissues, especially as the period of treatment increased. Although the sensitivity metric of the classifier presented higher values in the untreated tissues compared with the treated ones, the specificity and accuracy metrics demonstrated higher values for the treated tissues compared with the untreated ones. Conclusions: The biocompatible glycerol solution is a potential optical clearance agent in THz imaging while keeping the histopathology imaging intact. The SAM technique provided a good classification of cancerous tissues despite the small amount of cancer in the training data (only 7%). The SAM exponential kernel and EMD presented classification accuracy of ∼ 80 % to 85% compared with linear and polynomial kernels that provided accuracy ranging from 70% to 80%. Overall, glycerol treatment provides a potential improvement in cancer classification in freshly excised breast tumors.

Mammary tumors in Sprague Dawley rats induced by N-ethyl-N-nitrosourea for evaluating terahertz imaging of breast cancer.

Purpose: The objective of this study is to quantitatively evaluate terahertz (THz) imaging for differentiating cancerous from non-cancerous tissues in mammary tumors developed in response to injection of N-ethyl-N-nitrosourea (ENU) in Sprague Dawley rats. Approach: While previous studies have investigated the biology of mammary tumors of this model, the current work is the first study to employ an imaging modality to visualize these tumors. A pulsed THz imaging system is utilized to experimentally collect the time-domain reflection signals from each pixel of the rat's excised tumor. A statistical segmentation algorithm based on the expectation-maximization (EM) classification method is implemented to quantitatively assess the obtained THz images. The model classification of cancer is reported in terms of the receiver operating characteristic (ROC) curves and the areas under the curves. Results: The obtained low-power microscopic images of 17 ENU-rat tumor sections exhibited the presence of healthy connective tissue adjacent to cancerous tissue. The results also demonstrated that high reflection THz signals were received from cancerous compared with non-cancerous tissues. Decent tumor classification was achieved using the EM method with values ranging from 83% to 96% in fresh tissues and 89% to 96% in formalin-fixed paraffin-embedded tissues. Conclusions: The proposed ENU breast tumor model of Sprague Dawley rats showed a potential to obtain cancerous tissues, such as human breast tumors, adjacent to healthy tissues. The implemented EM classification algorithm quantitatively demonstrated the ability of THz imaging in differentiating cancerous from non-cancerous tissues.

Terahertz Imaging and Characterization Protocol for Freshly Excised Breast Cancer Tumors.

This manuscript presents a protocol to handle, characterize, and image freshly excised human breast tumors using pulsed terahertz imaging and spectroscopy techniques. The protocol involves terahertz transmission mode at normal incidence and terahertz reflection mode at an oblique angle of 30°. The collected experimental data represent time domain pulses of the electric field. The terahertz electric field signal transmitted through a fixed point on the excised tissue is processed, through an analytical model, to extract the refractive index and absorption coefficient of the tissue. Utilizing a stepper motor scanner, the terahertz emitted pulse is reflected from each pixel on the tumor providing a planar image of different tissue regions. The image can be presented in time or frequency domain. Furthermore, the extracted data of the refractive index and absorption coefficient at each pixel are utilized to provide a tomographic terahertz image of the tumor. The protocol demonstrates clear differentiation between cancerous and healthy tissues. On the other hand, not adhering to the protocol can result in noisy or inaccurate images due to the presence of air bubbles and fluid remains on the tumor surface. The protocol provides a method for surgical margins assessment of breast tumors.

Breast Cancer Detection with Low-dimension Ordered Orthogonal Projection in Terahertz Imaging.

This paper proposes a new dimension reduction algorithm based on low-dimension ordered orthogonal projection (LOOP), which is used for cancer detection with terahertz (THz) images of freshly excised human breast cancer tissues. A THz image can be represented by a data cube with each pixel containing a high dimension spectrum vector covering several THz frequencies, where each frequency represents a different dimension in the vector. The proposed algorithm projects the high-dimension spectrum vector of each pixel within the THz image into a low-dimension subspace that contains the majority of the unique features embedded in the image. The low-dimension subspace is constructed by sequentially identifying its orthonormal basis vectors, such that each newly chosen basis vector represents the most unique information not contained by existing basis vectors. A multivariate Gaussian mixture model is used to represent the statistical distributions of the low-dimension feature vectors obtained from the proposed dimension reduction algorithm. The model parameters are iteratively learned by using unsupervised learning methods such as Markov chain Monte Carlo or expectation maximization, and the results are used to classify the various regions within a tumor sample. Experiment results demonstrate that the proposed method achieves apparent performance improvement in human breast cancer tissue over existing approaches such as one-dimension Markov chain Monte Carlo. The results confirm that the dimension reduction algorithm presented in this paper is a promising technique for breast cancer detection with THz images, and the classification results present a good correlation with respect to the histopathology results of the analyzed samples.

Terahertz tomographic imaging of freshly excised human breast tissues.

Terahertz imaging and spectroscopy characterization of freshly excised breast cancer tumors are presented in the range 0.15 to 3.5 THz. Cancerous breast tissues were obtained from partial or full removal of malignant tumors while healthy breast tissues were obtained from breast reduction surgeries. The reflection spectroscopy to obtain the refractive index and absorption coefficient is performed on experimental data at each pixel of the tissue, forming tomographic images. The transmission spectroscopy of the refractive index and absorption coefficient are retrieved from experimental data at few tissue points. The average refractive index and absorption coefficients for cancer, fat, and collagen tissue regions are compared between transmission and reflection modes. The reflection mode offers the advantage of retrieving the electrical properties across a significantly greater number of points without the need for sectioning or altering the freshly excised tissue as in the transmission mode. The terahertz spectral power images and the tomographic images demonstrated good qualitative comparison with pathology.

Cancer detection in excised breast tumors using terahertz imaging and spectroscopy.

Terahertz imaging and spectroscopy has demonstrated a potential for differentiating tissue types of excised breast cancer tumors. Pulsed terahertz technology provides a broadband frequency range from 0.1 THz to 4 THz for detecting cancerous tissue. Tumor tissue types of interest include cancer typically manifested as infiltrating ductal or lobular carcinomas, fibro-glandular (healthy connective tissues) and fat. In this work, images of breast tumors excised from human and animal models are reviewed. In addition to alternate fresh tissues, breast cancer tissue phantoms are developed to further evaluate terahertz imaging and the potential use of contrast agents. Terahertz results are successfully validated with pathology images, showing strong differentiation between cancerous and healthy tissues for all freshly excised tissues and types. The advantages, challenges and limitations of THz imaging of breast cancer are discussed.

Pulsed terahertz imaging of breast cancer in freshly excised murine tumors.

This paper investigates terahertz (THz) imaging and classification of freshly excised murine xenograft breast cancer tumors. These tumors are grown via injection of E0771 breast adenocarcinoma cells into the flank of mice maintained on high-fat diet. Within 1 h of excision, the tumor and adjacent tissues are imaged using a pulsed THz system in the reflection mode. The THz images are classified using a statistical Bayesian mixture model with unsupervised and supervised approaches. Correlation with digitized pathology images is conducted using classification images assigned by a modal class decision rule. The corresponding receiver operating characteristic curves are obtained based on the classification results. A total of 13 tumor samples obtained from 9 tumors are investigated. The results show good correlation of THz images with pathology results in all samples of cancer and fat tissues. For tumor samples of cancer, fat, and muscle tissues, THz images show reasonable correlation with pathology where the primary challenge lies in the overlapping dielectric properties of cancer and muscle tissues. The use of a supervised regression approach shows improvement in the classification images although not consistently in all tissue regions. Advancing THz imaging of breast tumors from mice and the development of accurate statistical models will ultimately progress the technique for the assessment of human breast tumor margins.

Pulsed Terahertz Reflection Imaging of tumors in a spontaneous model of breast cancer.

We report the use of reflection-mode terahertz (THz) imaging in a transgenic mouse model of breast cancer. Unlike tumor xenografts that are grown from established cell lines, these tumors were spontaneously generated in the mammary fat pad of mice, and are a better representation of human breast cancer. THz imaging results from 7 tumors that recapitulate the compartmental complexity of breast cancer are presented here. Imaging was first performed on freshly excised tumors within an hour of excision and then repeated after fixation with formalin and paraffin. These THz images were then compared with histopathology to determine reflection-mode signals from specific regions within tumor. Our results demonstrate that the THz signal was consistently higher in cancerous tissue compared with fat, muscle, and fibrous tissue. Almost all tumors presented in this work demonstrated advanced stages where cancer infiltrated other tissues like fat and fibrous stroma. As the first known THz investigation in a transgenic model, these results hold promise for THz imaging at different stages of breast cancer.

Assessment of Terahertz Imaging for Excised Breast Cancer Tumors with Image Morphing.

This paper presents an image morphing algorithm for quantitative evaluation methodology of terahertz (THz) images of excised breast cancer tumors. Most current studies on the assessment of THz imaging rely on qualitative evaluation, and there is no established benchmark or procedure to quantify the THz imaging performance. The proposed morphing algorithm provides a tool to quantitatively align the THz image with the histopathology image. Freshly excised xenograft murine breast cancer tumors are imaged using the pulsed THz imaging and spectroscopy system in the reflection mode. Upon fixing the tumor tissue in formalin and embedding in paraffin, an FFPE tissue block is produced. A thin slice of the block is prepared for the pathology image while another THz reflection image is produced directly from the block. We developed an algorithm of mesh morphing using homography mapping of the histopathology image to adjust the alignment, shape, and resolution to match the external contour of the tissue in the THz image. Unlike conventional image morphing algorithms that rely on internal features of the source and target images, only the external contour of the tissue is used to avoid bias. Unsupervised Bayesian learning algorithm is applied to THz images to classify the tissue regions of cancer, fat, and muscles present in xenograft breast tumors. The results demonstrate that the proposed mesh morphing algorithm can provide more effective and accurate evaluation of THz imaging compared with existing algorithms. The results also showed that while THz images of FFPE tissue are highly in agreement with pathology images, challenges remain in assessing THz imaging of fresh tissue.

A Phantom Study of Terahertz Spectroscopy and Imaging of Micro- and Nano-diamonds and Nano-onions as Contrast Agents for Breast Cancer.

THz imaging is effective in distinguishing between cancerous, healthy, and fatty tissues in breast tumors, but a challenge remains in the contrast between cancerous and fibroglandular (healthy) tissues. This work investigates carbon-based nanoparticles as potential contrast agents for terahertz imaging of breast cancer. Microdiamonds, nanodiamonds, and nanometer-scale onion-like carbon are characterized with terahertz transmission spectroscopy in low-absorption backgrounds of polydimethylsiloxane or polyethylene. The refractive index and absorption coefficients are calculated based on the measured electric fields. Nanodiamonds show little effect on the terahertz signal, microdiamonds express resonance-like, size-dependent absorption peaks, and onion-like carbon provides a uniform increase in the optical properties even at low concentration. Due to its strong interaction with terahertz frequencies and ability to be activated for selective binding to cancer cells, onion-like carbon is implemented into engineered three-dimensional breast tumor models composed of phantom tissue mimicking infiltrating ductal carcinoma surrounded by a phantom mimicking healthy fibroglandular tissue. This model is imaged using the terahertz reflection mode to examine the effectiveness of contrast agents for differentiation between the two tissue types. In both spectroscopy and imaging, a 10% concentration of onion-like carbon shows the strongest impact on the terahertz signal and holds promise as a terahertz contrast agent.

Terahertz transmission vs reflection imaging and model-based characterization for excised breast carcinomas.

This work presents experimental and analytical comparison of terahertz transmission and reflection imaging modes for assessing breast carcinoma in excised paraffin-embedded human breast tissue. Modeling for both transmission and reflection imaging is developed. The refractive index and absorption coefficient of the tissue samples are obtained. The reflection measurements taken at the system's fixed oblique angle of 30° are shown to be a hybridization of TE and TM modes. The models are validated with transmission spectroscopy at fixed points on fresh bovine muscle and fat tissues. Images based on the calculated absorption coefficient and index of refraction of bovine tissue are successfully compared with the terahertz magnitude and phase measured in the reflection mode. The validated techniques are extended to 20 and 30 µm slices of fixed human lobular carcinoma and infiltrating ductal carcinoma mounted on polystyrene microscope slides in order to investigate the terahertz differentiation of the carcinoma with non-cancerous tissue. Both transmission and reflection imaging show clear differentiation in carcinoma versus healthy tissue. However, when using the reflection mode, in the calculation of the thin tissue properties, the absorption is shown to be sensitive to small phase variations that arise due to deviations in slide and tissue thickness and non-ideal tissue adhesion. On the other hand, the results show that the transmission mode is much less sensitive to these phase variations. The results also demonstrate that reflection imaging provides higher resolution and more clear margins between cancerous and fibroglandular regions, cancerous and fatty regions, and fibroglandular and fatty tissue regions. In addition, more features consistent with high power pathology images are exhibited in the reflection mode images.

Biopotential signals of breast cancer versus tumor types and proliferation stages.

Clinical studies have shown compelling data of elevated biopotential signals recorded noninvasively from the breasts of women with breast cancer. While these data are compelling and show a strong potential for use in the noninvasive early detection of breast cancer, there remains significant knowledge gaps which must be addressed before this technology can be routinely used for breast cancer detection. A diffusion-drift model is developed to study the spatial and temporal characteristics of the biopotential signals of breast tumors taking into account the morphology and cell division stages. The electric signals of the most common tumor types-papillary, compact, and comedo-are also considered. The largest biopotential signal is observed from the compact tumor, while the smallest signal is observed from the papillary type. The results also show an increase in the time duration of the generated biopotential signals when cancer cells start their transitions at different time instants. The spatial and temporal variations of the biopotential signals are correlated with the tumor pattern which can have important implications for breast cancer detection.

Review of electromagnetic techniques for breast cancer detection.

Breast cancer is anticipated to be responsible for almost 40,000 deaths in the USA in 2011. The current clinical detection techniques suffer from limitations which motivated researchers to investigate alternative modalities for the early detection of breast cancer. This paper focuses on reviewing the main electromagnetic techniques for breast cancer detection. More specifically, this work reviews the cutting edge research in microwave imaging, electrical impedance tomography, diffuse optical tomography, microwave radiometry, biomagnetic detection, biopotential detection, and magnetic resonance imaging (MRI). The goal of this paper is to provide biomedical researchers with an in-depth review that includes all main electromagnetic techniques in the literature and the latest progress in each of these techniques.

Modeling biopotential signals and current densities of multiple breast cancerous cells.

This study presents a model to simulate the electrophysiological activities of multiple Michigan Cancer Foundation-7 (MCF-7) cells, the most studied breast cancer cell line. The intercellular spacing of MCF-7 cells is estimated using the effective diffusion coefficient. Nonuniform finite-difference discretization is implemented to accommodate for the contrast in size between the intercellular spacing and the cell dimension. The model computes the amplitude and the spatiotemporal patterns of biopotentials and electric current densities at different cell division stages. The results show that the biopotentials increase proportionally to the number of cells, especially when all cells are in the hyperpolarization stage. Also, the results show significant electric current density in the intercellular gap between the cells.

Modeling electrical activities of a growing breast cancerous cell based on a semiconductor approach.

A two dimensional diffusion-drift model is developed to simulate the electrical activities of a breast cancerous cell during the hyperpolarization which occurs at the G1/S transition. The model focuses on calculating the temporal and the spatial patterns of the electric current densities and biopotentials generated at the cell boundary and its surroundings. Different durations for the hyperpolarization phase were studied. The results show that the electric signals generated in the tumor cell's surroundings increase as the duration of the hyperpolarization phase decreases or when the transition becomes faster.

Diffusion-drift modeling of a growing breast cancerous cell.

This paper presents a 2-D model to calculate the electric current densities and the biopotential differences generated due to a breast cancerous cell during the hyperpolarization of the G1/synthesis (G1/S) transition. The proposed model is based on semiconductor diffusion-drift analysis, and aims to understand the biophysics associated with growing breast cancerous cells. The effect of the duration of the G1/S transition, and the diffusivity and the mobility of the cancerous cell boundary is investigated. The results show that shorter G1/S transition durations, and higher diffusivity and mobility at the cell boundary provide higher magnitude of the electric signals.

Computational model of ductal carcinoma in situ: the effects of contact inhibition on pattern formation.

The computational model presented in this paper focuses on modeling ductal carcinoma in situ (DCIS), which is the most commonly detected preinvasive form of breast cancer. The model aims to understand the biological mechanisms and resultant growth dynamics of DCIS. The cellular automaton model based on observed phenotypic characteristics of DCIS emphasize the important role of contact inhibition on lesion pattern formation. Computer simulations resembled the cribriform, micropapillary, solid, and comedo patterns of DCIS. The model has led to insights about the progression of the preinvasive disease such as possible explanations for coexisting micropapillary and cribriform patterns commonly found through histological analyses.

Spherical harmonics microwave algorithm for shape and location reconstruction of breast cancer tumor.

A reconstruction algorithm to simultaneously estimate the shape and location of three-dimensional breast cancer tumor is presented and its utility is analyzed. The approach is based on a spherical harmonic decomposition to capture the shape of the tumor. We combine a gradient descent optimization method with a direct electromagnetic solver to determine the coefficients in the harmonic expansion as well as the coordinates of the center of the tumor. The results demonstrate the potential advantage of collecting data using a multiple-view/tomographic-type strategy. We show how the order of the harmonic expansion must be increased to capture increasingly "irregularly" shaped tumors and explore the resulting increase in the central processing unit (CPU) time required by the algorithm. Our approach shows accurate reconstruction of the tumor image regardless of the source polarization. This work demonstrates the promise of the algorithm when used on data corrupted with Gaussian noise and when perfect knowledge of the tumor electrical properties is not available.

Resonant spectra of malignant breast cancer tumors using the three-dimensional electromagnetic fast multipole model.

This paper presents an intensive numerical study of the resonance scattering of malignant breast cancer tumors. The three-dimensional electromagnetic model, based on the equivalence theorem, is used to obtain induced electric and magnetic currents on breast and tumor surfaces. The results show that the nonspherical malignant tumor can be characterized, based on its spectra, regardless of orientation, incident polarization, or incident or scattered directions. The spectra of the tumor depend solely upon its physical characteristics (i.e., shape and electrical properties); however, their locations are not functions of the depth of the tumor beneath the breast surface. This paper can be a guide in the selection of the frequency range at which the tumor resonates to produce the maximum signature at the receiver.