Search

Purpose In low- to middle-income countries (LMICs), most breast cancers present as palpable lumps; however, most palpable lumps are benign. We have developed artificial intelligence–based computer-assisted diagnosis (CADx) for an existing low-cost portable ultrasound system to triage which lumps need further evaluation and which are clearly benign. This pilot study was conducted to demonstrate that this approach can be successfully used by minimally trained health care workers in an LMIC country. Patients and Methods We recruited and trained three nonradiologist health care workers to participate in an institutional review board–approved, Health Insurance Portability and Accountability Act–compliant pilot study in Jalisco, Mexico, to determine whether they could use portable ultrasound (GE Vscan Dual Probe) to acquire images of palpable breast lumps of adequate quality for accurate computer analysis. Images from 32 women with 32 breast masses were then analyzed with a triage-CADx system, generating an output of benign or suspicious (biopsy recommended). Triage-CADx outputs were compared with radiologist readings. Results The nonradiologists were able to acquire adequate images. Triage by the CADx software was as accurate as assessment by specialist radiologists, with two (100%) of two cancers considered suspicious and 30 (100%) of 30 benign lesions classified as benign. Conclusion A portable ultrasound system with CADx software can be successfully used by first-level health care workers to triage palpable breast lumps. These results open up the possibility of implementing practical, cost-effective triage of palpable breast lumps, ensuring that scarce resources can be dedicated to suspicious lesions requiring further workup.

The use of photographs of the screen of displayed medical images is explored to circumvent the challenges involved in transferring images between sites. The photographs can be conveniently taken with a smartphone and analyzed remotely by either human or AI experts. An autoencoder preprocessor is shown to improve the performance for human experts. The AI performance provided by photographs is shown to be statistically equivalent to using the original DICOM images. The autoencoder preprocessor increases the PSNR by 15 dB or greater and provides an AUC that is statistically equivalent to using the original DICOM images. The photo approach is an alternative to IHE-based teleradiology applications while avoiding the problems inherit in navigating the proprietary and security barriers that limit DICOM communication between PACS in practice., Comment: 4 pages, 9 figures , IEEE SPMB19 conference submission

Medical Ultrasonography is a valuable imaging technology for medical diagnostics and, more recently, as a screening alternative to mammography for women with dense breasts. However, ultrasound imaging within the contexts of both diagnostic and screening mammography suffers from inter-operator and intra-operator variability. Consequently, there is a broad distribution of performance profiles, even for radiologists of similar training. Typically, these profiles tend to err on the side of caution, preferring false positive errors to false negative errors. While this approach may lead to a higher Cancer Detection Rate (CDR), it also lowers the Positive Predictive Value (PPV3) of performed biopsies. A lower PPV 3 translates to an increase in benign biopsies, the annual cost of which are estimated to be on the order of $1 – $3 billion USD (not including pathological workups). And, of course, there is the immeasurable cost of pain, worry, and suffering borne by women undergoing these potentially unnecessary procedures. In this paper, we evaluate the ability of the ClearView cCAD algorithms to increase overall performance and reduce the inter-operator variance on a set of imaged lesions. The cCAD system provides an automated assessment of some ACR BI-RADs criteria and calculates a preliminary BI-RADs assessment, given as BI-RADS categorical bucket (1–3) or (4–5). Through the evaluation of 1300 breast lesion images, 3 MQSA certified radiologists were asked to determine both a Likelihood of Malignancy (LoM) and a BI-RADs assessment, from which their ROC curve AUC as well as PPV 3 could be calculated. The cCAD system was also evaluated, on the same set of lesions, by a similar set of metrics. From this analysis we have been able to show that the cCAD system outperforms radiologists at all operating points within the scope of this study design. Furthermore, we've shown that through simple fusion schemes we are able to increase performance beyond that of either the cCAD system or the radiologist alone by all typically tracked quality metrics, and significantly reduce inter-operator variance.

We propose an automated breast cancer triage CAD system using machine vision on low-cost, portable ultrasound imaging devices. We demonstrate that the triage CAD software can effectively analyze images captured by minimally-trained operators and output one of three assessments — benign, probably benign (6-month follow-up recommended) and suspicious (biopsy recommended). This system opens up the possibility of offering practical, cost-effective breast cancer diagnosis for symptomatic women in economically developing countries.

Speaker recognition refers to the concept of recognizing a speaker by his/her voice or speech samples. Some of the important applications of speaker recognition include customer verification for bank transactions, access to bank accounts through telephones, control on the use of credit cards, and for security purposes in the army, navy and airforce. This paper is purely a tutorial that presents a review of the classifier based methods used for speaker recognition. Both unsupervised and supervised classifiers are described. In addition, practical approaches that utilize diversity, redundancy and fusion strategies are discussed with the aim of improving performance.

Background: According to GLOBOCAN, 1.7 million women were diagnosed with breast cancer and 445,000 died from the disease in 2012. Breast cancer is the most common cause of cancer death among women worldwide and the numbers are disproportionately high for women in developing countries. Furthermore, breast cancer is more common in young women (age 44 and under) in these countries than in industrialized countries. Shifts in lifestyle typical of industrialized societies are thought to be causing a rise in breast cancer incidence while technical advances in detection, diagnosis, and treatment are not reaching these areas of the world. In developing countries, breast cancer commonly presents in women as a palpable mass and a low-cost solution for breast cancer triage would be beneficial. Ultrasound has been shown to perform very well on palpable masses, is non-ionizing, and affordable, portable systems are available. In the ACRIN 6666 clinical study of asymptomatic women, adding screening ultrasound to mammography significantly increased breast cancer detection in women with elevated cancer risk and dense breasts, and more invasive cancers were seen by ultrasound alone than by mammography alone. Method: We propose a novel computer-aided diagnosis (CADx) tool for breast cancer triage using a low-cost ultrasound imaging device that will automatically distinguish among (1) suspicious lesions to be sent for biopsy, (2) benign findings that need no additional follow-up and (3) probably benign findings with recommended 6-month follow-up. A validation study at USC Norris Cancer Center and Los Angeles Biomedical Research Institute at Harbor-UCLA Medical Center was performed in order to assess the performance of the triage software compared to the standard-of-care and biopsy-confirmed cancers and false positives. The Study machine was the GE Logiq e ultrasound imaging device, which is FDA-approved for small body parts. Lesions that were assessed as BI-RADS 2 by radiologists were determined to be benign while any lesion assessed as BI-RADS 4a or higher was sent for biopsy. This study cohort included women 61 years of age or younger with a palpable mass. The triage software was compared to the performance by radiologists using conventional standard-of-care for palpable masses. Results: The original milestone for this study was to achieve a sensitivity of 85% while reducing the number of benign lesions going to biopsy by 40%. The validation study showed an actual sensitivity of 100% while reducing the number of benign lesions going to biopsy by 69%. To date, the images collected during the validation study include 152 lesions in total, 22 biopsy-proven cancers and 130 benign lesions; 63 benign lesions were categorized as BI-RADS 4a or higher by a radiologist and sent for biopsy. Conclusions: The validation study showed that adding automated breast cancer triage software to a low-cost portable ultrasound imaging device is a realistic approach for regions of the world with limited access to highly trained radiologists and diagnostic equipment. Funding: NIH Grant #UH2EB019889. Citation Format: Love SM, Berg WA, Podilchuk C, Hovanessian-Larsen LJ, Dauphine C, Jairaj A, Barinov L, Hulbert W, Cen S, Eshraghi L, Mammone R. Automated, low-cost palpable breast lump triage for economically-developing countries [abstract]. In: Proceedings of the 2016 San Antonio Breast Cancer Symposium; 2016 Dec 6-10; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2017;77(4 Suppl):Abstract nr PD3-01.

Recently, a new speckle noise reduction and contrast enhancement technique has been introduced that is motivated by the research in compressive sampling or sensing. Compressive sampling is based on the principle that a sparse signal such as ultrasound can be fully recovered when sampled below the Nyquist rate. This allows for a new noise reduction technique that preserves the high frequency and fine details while reducing the effects of speckle noise. This method improves the overall perceptual quality of the image for visualization and diagnosis by the radiologist. This paper examines how the improvement in SNR makes the method suitable as a preprocessor to improve a computer aided detection (CAD) system for breast cancer detection. Classical performance metrics such as false positive rates, false negative rates and receiver operator curves will be used to show the benefits of this approach. Initial experiments look promising for microcalcification detection, where the new method yields a false negative rate of 20 percent at a false positive rate of 0.5 percent while the traditional speckle reduction techniques yield a false negative rate of 60 percent at a false positive rate of 0.5 percent.

Medical Ultrasonography is a valuable imaging technology for medical diagnostics and to guide interventional procedures. However, ultrasound imaging suffers from speckle noise, an inherent characteristic of all coherent imaging techniques due to the presence of sub-resolution scatterers. Speckle noise produces a reduction in contrast resolution which is responsible for the overall lower effective resolution of ultrasound compared to x-ray or MRI imaging. In the case of breast imaging, ultrasound speckle can mask small details such as low contrast tumors or microcalcifications, which may be an early indication of breast cancer. This limitation prevents ultrasound from displacing mammography as the gold standard for breast cancer screening. Traditional speckle reduction techniques attempt to remove speckle noise while preserving edges and other important features but there is always a tradeoff between removing the speckle noise and blurring tissue structure and details. We introduce a novel speckle reduction and contrast enhancement method for ultrasound imaging that is motivated by the fundamental ideas behind compressive sampling. We also introduce a way to estimate instantaneous SNR in order to identify the areas that are mostly signal from the areas that are mostly noise in order to preserve the signal while suppressing the noise. We have shown improvements in SNR on the order of 12dB in the lab and improved visualization of clinical data.

A common problem in speaker identification systems is that a mismatch in the training and testing conditions sacrifices much performance. We attempt to alleviate this problem by proposing new features that show less variation when speech is corrupted by convolutional noise (channel) and/or additive noise. The conventional feature used is the linear predictive (LP) cepstrum that is derived from an all-pole transfer function which, in turn, achieves a good approximation to the spectral envelope of the speech. A different cepstral feature based on a pole-zero function (called the adaptive component weighted or ACW cepstrum) was previously introduced. We propose four additional new cepstral features based on pole-zero transfer functions. One is an alternative way of doing adaptive component weighting and is called the ACW2 cepstrum. Two others (known as the PFL1 cepstrum and the PFL2 cepstrum) are based on a pole-zero postfilter used in speech enhancement. Finally, an autoregressive moving-average (ARMA) analysis of speech results in a pole-zero transfer function describing the spectral envelope. The cepstrum of this transfer function is the feature. Experiments involving a closed set, text-independent and vector quantizer based speaker identification system are done to compare the various features. The TIMIT and King databases are used. The ACW and PFL1 features are the preferred features, since they do as well or better than the LP cepstrum for all the test conditions. The corresponding spectra show a clear emphasis of the formants and no spectral tilt.

Medical Ultrasonography is a valuable imaging technology for medical diagnostics and to guide interventional procedures. Ultrasound imaging is particularly useful in breast cancer detection and diagnosis for women with dense breast tissue where traditional mammography may fail to detect suspicious areas. However, ultrasound imaging suffers from speckle noise, an inherent characteristic of all coherent imaging techniques due to the presence of sub-resolution scatterers. Speckle noise produces a reduction in contrast resolution which is responsible for the overall lower effective resolution of ultrasound compared to x-ray or MRI imaging. In the case of breast imaging, ultrasound speckle can mask small details such as low contrast tumors or micro-calcifications, which may be an early indication of breast cancer. This limitation prevents ultrasound from displacing mammography as the gold standard for breast cancer screening. In conventional pulsed ultrasound imaging systems, de-noising techniques are used to minimize the effect of speckle noise. However, research shows that there is a tradeoff between the effectiveness of speckle reduction techniques and image resolution. We introduce stepped-frequency continuous wave ultrasound imaging which provides a framework where speckle reduction techniques are particularly effective, resulting in higher quality images with an improved SNR and significantly lower speckle noise while maintaining the spatial resolution of the original scan so that small lesions of interest are visible to the radiologist.

Various linear predictive (LP) analysis methods are studied and compared from the points of view of robustness to noise and of application to speaker identification. The key to the success of the LP techniques is in separating the vocal tract information from the pitch information present in a speech signal even under noisy conditions. In addition to considering the conventional, one-shot weighted least-squares methods, the authors propose three other approaches with the above point as a motivation. The first is an iterative approach that leads to the weighted least absolute value solution. The second is an extension of the one-shot least-squares approach and achieves an iterative update of the weights. The update is a function of the residual and is based on minimizing a Mahalanobis distance. Third, the weighted total least-squares formulation is considered. A study of the deviations in the LP parameters is done when noise (white Gaussian and impulsive) is added to the speech. It is revealed that the most robust method depends on the type of noise. Closed-set speaker identification experiments with 20 speakers are conducted using a vector quantizer classifier trained on clean speech. The relative performance of the various LP approaches depends on the type of speech material used for testing. >

An evaluation of various classifiers for text-independent speaker recognition is presented. In addition, a new classifier is examined for this application. The new classifier is called the modified neural tree network (MNTN). The MNTN is a hierarchical classifier that combines the properties of decision trees and feedforward neural networks. The MNTN differs from the standard NTN in both the new learning rule used and the pruning criteria. The MNTN is evaluated for several speaker recognition experiments. These include closed- and open-set speaker identification and speaker verification. The database used is a subset of the TIMIT database consisting of 38 speakers from the same dialect region. The MNTN is compared with nearest neighbor classifiers, full-search, and tree-structured vector quantization (VQ) classifiers, multilayer perceptrons (MLPs), and decision trees. For closed-set speaker identification experiments, the full-search VQ classifier and MNTN demonstrate comparable performance. Both methods perform significantly better than the other classifiers for this task. The MNTN and full-search VQ classifiers are also compared for several speaker verification and open-set speaker-identification experiments. The MNTN is found to perform better than full-search VQ classifiers for both of these applications. In addition to matching or exceeding the performance of the VQ classifier for these applications, the MNTN also provides a logarithmic saving for retrieval. >

A new set of features is introduced that has been found to improve the performance of automatic speaker identification systems, The new set of features is referred to as the adaptive component weighting (ACW) cepstral coefficients. The new features emphasize the formant structure of the speech spectrum while attenuating the broad-bandwidth spectral components. The attenuated components correspond to the variations in spectral tilt of transmission and recording environment, and other characteristics that are irrelevant to speaker identification. The resulting ACW spectrum introduces zeros into the usual all-pole linear prediction (LP) spectrum. This is equivalent to applying a finite impulse response (FIR) filter that normalizes the narrow-band modes of the spectrum. Unlike existing fixed cepstral weighting schemes, the ACW cepstrum provides an adaptively weighted version of the LP cepstrum. The adaptation results in deemphasizing the irrelevant variations of the LP cepstral coefficients on a frame-by-frame basis. The ACW features are evaluated for text-independent speaker identification and are shown to yield improved performance. >