Data sets with relatively few observations (cases) in medical research are common, especially if the data are expensive or difficult to collect. Such small sample sizes usually do not provide enough information for computer models to learn data patterns well enough for good prediction and generalization. As a model that may be able to maintain good classification performance in the presence of limited data, we used decision fusion. In this study, we investigated the effect of sample size on the generalization ability of both linear discriminant analysis (LDA) and decision fusion. Subsets of large data sets were selected by a bootstrap sampling method, which allowed us to estimate the mean and standard deviation of the classification performance as a function of data set size. We applied the models to two breast cancer data sets and compared the models using receiver operating characteristic (ROC) analysis. For the more challenging calcification data set, decision fusion reached its maximum classification performance of AUC = 0.80±0.04 at 50 samples and pAUC = 0.34±0.05 at 100 samples. The LDA reached a lower performance and required many more cases, with a maximum of AUC = 0.68±0.04 and pAUC = 0.12±0.05 at 450 samples. For the mass data set, the two classifiers had more similar performance, with AUC = 0.92±0.02 and pAUC = 0.48±0.02 at 50 samples for decision fusion and AUC = 0.92±0.03 and pAUC = 0.55±0.04 at 500 samples for the LDA.
Previous research has shown benign and cancerous tissues to have different chemical make-ups. To measure the elemental concentration of biological samples noninvasively, we used neutron stimulated emission computed tomography (NSECT). When an incident neutron scatters inelastically from an atomic nucleus, it emits characteristic gamma energies, allowing for measurement of the elemental concentration of biological samples. Thus NSECT has the potential to be a method for precancerous tissue detection. In Monte Carlo simulations, we bombarded both a benign and a malignant human breast with 50 million neutrons. The resulting photon spectra were blurred to model the detector resolutions and then analyzed for peak detection. This simulation study analyzed the characteristic spectra using three detectors of different resolutions: a High-Purity Germanium (HPGe) semiconductor, a Bismuth Germanate (BGO) scintillator, and a Sodium Iodide (NaI) scintillator. The effective energy resolutions of these detectors are 0.1%, 7%, and 12%, respectively. The detectability of element peaks in the breast model was greatly reduced when the blur increased from just 0.1% to 7%. These initial experiments are valuable in choosing optimal detectors for peak detection in further NSECT studies and indicate that high-resolution detectors, such as HPGe, are required for using spectral peak analysis for breast cancer prediction.
KEYWORDS: Sensors, Modulation transfer functions, Digital mammography, Prototyping, Molybdenum, Tungsten, Aluminum, Calibration, Quantum efficiency, Signal to noise ratio
This study evaluated the physical performance of a selenium-based direct full-field digital mammography prototype detector (Siemens Mammomat NovationDR), including the pixel value vs. exposure linearity, the modulation transfer function (MTF), the normalized noise power spectrum (NNPS), and the detective quantum efficiency (DQE). The current detector is the same model which received an approvable letter from FDA for release to the US market. The results of the current prototype are compared to those of an earlier prototype. Two IEC standard beam qualities (RQA-M2: Mo/Mo, 28 kVp, 2 mm Al; RQA-M4: Mo/Mo, 35 kVp, 2 mm Al) and two additional beam qualities (MW2: W/Rh, 28 kVp, 2 mm Al; MW4: W/Rh, 35 kVp, 2 mm Al) were investigated. To calculate the modulation transfer function (MTF), a 0.1 mm Pt-Ir edge was imaged at each beam quality. Detector pixel values responded linearly against exposure values (R2 0.999). As before, above 6 cycles/mm Mo/Mo MTF was slightly higher along the chest-nipple axis compared to the left-right axis. MTF was comparable to the previously reported prototype, with slightly reduced resolution. The DQE peaks ranged from 0.71 for 3.31 μC/kg (12.83 mR) to 0.4 for 0.48 μC/kg (1.86 mR) at 1.75 cycles/mm for Mo/Mo at 28 kVp. The DQE range for W/Rh at 28 kVP was 0.81 at 2.03 μC/kg (7.87 mR) to 0.50 at 0.50 μC/kg (1.94 mR) at 1 cycle/mm. NNPS tended to increase with greater exposures, while all exposures had a significant low-frequency component. Bloom and detector edge artifacts observed previously were no longer present in this prototype. The new detector shows marked noise improvement, with slightly reduced resolution. There remain artifacts due to imperfect gain calibration, but at a reduced magnitude compared to a prototype detector.
A prototype breast tomosynthesis system has been developed, allowing a total angular view of ±25°. The detector used in this system is an amorphous selenium direct-conversion digital flat-panel detector suitable for digital tomosynthesis. The system is equipped with various readout sequences to allow the investigation of different tomosynthetic data acquisition modes. In this paper, we will present basic physical properties -- such as MTF, NPS, and DQE -- measured for the full resolution mode and a binned readout mode of the detector. From the measured projections, slices are reconstructed employing a special version of filtered backprojection algorithm. In a phantom study, we compare binned and full resolution acquisition modes with respect to image quality. Under the condition of same dose, we investigate the impact of the number of views on artifacts. Finally, we show tomosynthesis images reconstructed from first clinical data.
We developed an ensemble classifier for the task of computer-aided diagnosis of breast microcalcification clusters,which are very challenging to characterize for radiologists and computer models alike. The purpose of this study is to help radiologists identify whether suspicious calcification clusters are benign vs. malignant, such that they may potentially recommend fewer unnecessary biopsies for actually benign lesions. The data consists of mammographic features extracted by automated image processing algorithms as well as manually interpreted by radiologists according to a standardized lexicon. We used 292 cases from a publicly available mammography database. From each cases, we extracted 22 image processing features pertaining to lesion morphology, 5 radiologist features also pertaining to morphology, and the patient age. Linear discriminant analysis (LDA) models were designed using each of the three data types. Each local model performed poorly; the best was one based upon image processing features which yielded ROC area index AZ of 0.59 ± 0.03 and partial AZ above 90% sensitivity of 0.08 ± 0.03. We then developed ensemble models using different combinations of those data types, and these models all improved performance compared to the local models. The final ensemble model was based upon 5 features selected by stepwise LDA from all 28 available features. This ensemble performed with AZ of 0.69 ± 0.03 and partial AZ of 0.21 ± 0.04, which was statistically significantly better than the model based on the image processing features alone (p<0.001 and p=0.01 for full and partial AZ respectively). This demonstrated the value of the radiologist-extracted features as a source of information for this task. It also suggested there is potential for improved performance using this ensemble classifier approach to combine different sources of currently available data.
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.