Stella Urban

Recent developments in tissue-type imaging (TTI) for planning and monitoring treatment of prostate cancer.

Because current methods of imaging prostate cancer are inadequate, biopsies cannot be effectively guided and treatment cannot be effectively planned and targeted. Therefore, our research is aimed at ultrasonically characterizing cancerous prostate tissue so that we can image it more effectively and thereby provide improved means of detecting, treating and monitoring prostate cancer. We base our characterization methods on spectrum analysis of radiofrequency (rf) echo signals combined with clinical variables such as prostate-specific antigen (PSA). Tissue typing using these parameters is performed by artificial neural networks. We employed and evaluated different approaches to data partitioning into training, validation, and test sets and different neural network configuration options. In this manner, we sought to determine what neural network configuration is optimal for these data and also to assess possible bias that might exist due to correlations among different data entries among the data for a given patient. The classification efficacy of each neural network configuration and data-partitioning method was measured using relative-operating-characteristic (ROC) methods. Neural network classification based on spectral parameters combined with clinical data generally produced ROC-curve areas of 0.80 compared to curve areas of 0.64 for conventional transrectal ultrasound imaging combined with clinical data. We then used the optimal neural network configuration to generate lookup tables that translate local spectral parameter values and global clinical-variable values into pixel values in tissue-type images (TTIs). TTIs continue to show cancerous regions successfully, and may prove to be particularly useful clinically in combination with other ultrasonic and nonultrasonic methods, e.g., magnetic-resonance spectroscopy.

Spectrum-Analysis and Neural Networks for Imaging to Detect and Treat Prostate Cancer

Conventional B-mode ultrasound currently is the standard means of imaging the prostate for guiding prostate biopsies and planning brachytherapy to treat prostate cancer. Yet B-mode images do not adequately display cancerous lesions of the prostate. Ultrasonic tissue-type imaging based on spectrum analysis of radiofrequency (rf) echo signals has shown promise for overcoming the limitations of B-mode imaging for visualizing prostate tumors. This method of tissue-type imaging utilizes nonlinear classifiers, such as neural networks, to classify tissue based on values of spectral parameter and clinical variables. Two- and three-dimensional images based on these methods demonstrate potential for guiding prostate biopsies and targeting radiotherapy of prostate cancer. Two-dimensional images are being generated in real time in ultrasound scanners used for real-time biopsy guidance and have been incorporated into commercial dosimetry software used for brachytherapy planning. Three-dimensional renderings show promise for depicting locations and volumes of cancer foci for disease evaluation to assist staging and treatment planning, and potentially for registration or fusion with CT images for targeting external-beam radiotherapy.

ROLE OF ADVANCED 2 AND 3-DIMENSIONAL ULTRASOUND FOR DETECTING PROSTATE CANCER

PURPOSE We explored the clinical usefulness of spectrum analysis and neural networks for classifying prostate tissue and identifying prostate cancer in patients undergoing transrectal ultrasound for diagnostic or therapeutic reasons. MATERIALS AND METHODS Data on a cohort of 215 patients who underwent transrectal ultrasound guided prostate biopsies at Memorial-Sloan Kettering Cancer Center, New York, New York were included in this study. Radio frequency data necessary for 2 and 3-dimensional (D) computer reconstruction of the prostate were digitally recorded at transrectal ultrasound and prostate biopsy. The data were spectrally processed and 2-D tissue typing images were generated based on a pre-trained neural network classification. We used manually masked 2-D tissue images as building blocks for generating 3-D tissue images and the images were tissue type color coded using custom software. Radio frequency data on the study cohort were analyzed for cancer probability using the data set pre-trained by neural network methods and compared with conventional B-mode imaging. ROC curves were generated for the 2 methods using biopsy results as the gold standard. RESULTS The mean area under the ROC curve plus or minus SEM for detecting prostate cancer for the conventional B-mode and neural network methods was 0.66 +/- 0.03 and 0.80 +/- 0.05, respectively. Sensitivity and specificity for detecting prostate cancer by the neural network method were significantly increased compared with conventional B-mode imaging. In addition, the 2 and 3-D prostate images provided excellent visual identification of areas with a higher likelihood of cancer. CONCLUSIONS Spectrum analysis could significantly improve the detection and evaluation of prostate cancer. Routine real-time application of spectrum analysis may significantly decrease the number of false-negative biopsies and improve the detection of prostate cancer at transrectal ultrasound guided prostate biopsy. It may also provide improved identification of prostate cancer foci during therapeutic intervention, such as brachytherapy, external beam radiotherapy or cryotherapy. In addition, 2 and 3-D images with prostate cancer foci specifically identified can help surgical planning and may in the distant future be an additional reliable noninvasive method of selecting patients for prostate biopsy.

Ultrasonic spectrum-analysis and neural-network classification as a basis for ultrasonic imaging to target brachytherapy of prostate cancer

Conventional B-mode ultrasound is the standard means of imaging the prostate for guiding prostate biopsies and planning brachytherapy of prostate cancer. Yet B-mode images do not allow adequate visualization of cancerous lesions of the prostate. Ultrasonic tissue-typing imaging based on spectrum analysis of radiofrequency echo signals has shown promise for overcoming the limitations of B-mode imaging for visualizing prostate tumors. Tissue typing based on radiofrequency spectrum analysis uses nonlinear methods, such as neural networks, to classify tissue by using spectral-parameter and clinical-variable values. Two- and three-dimensional images based on these methods show potential for improving the guidance of prostate biopsies and the targeting of radiotherapy of prostate cancer. Two-dimensional images have been imported into instrumentation for real-time biopsy guidance and into commercial dose-planning software for brachytherapy planning. Three-dimensional renderings seem to be capable of depicting locations and volumes of cancer foci.

Application of spectrum analysis and neural-network classification to imaging for targeting and monitoring treatment of prostate cancer

Conventional B-mode ultrasound is the standard means of imaging the prostate for guiding prostate biopsies and planning brachytherapy of prostate cancer. Yet no imaging modality, including B-mode images, reliably shows cancerous lesions of the prostate. Tissue-typing imaging based on spectrum analysis of ultrasonic radio-frequency (RF) echo signals may be able to overcome the limitations of conventional imaging modalities for visualizing prostate tumors. Such tissue typing utilizes neural-networks, to classify tissue based on spectral-parameter and clinical-variable values. Tissue-type images based on these methods are intended to improve guidance of prostate biopsies and targeting of radiotherapy of prostate cancer. Two-dimensional images have been imported into instrumentation for real-time biopsy guidance and into commercial dose-planning software for brachytherapy planning. Three-dimensional renderings show locations and volumes of cancer foci.

Prostate imaging based on rf spectrum analysis and nonlinear classifiers for guiding biopsies and targeting radiotherapy

Conventional B-mode ultrasound is the standard means of imaging the prostate for guiding prostate biopsies and planning radiotherapy (i.e., brachytherapy and external-beam radiation) of prostate cancer (CaP). Yet B-mode images essentially do not allow visualization of cancerous lesions of the prostate. Ultrasonic tissue-typing imaging based on spectrum analysis of radio-frequency (RF) echo signals has shown promise for overcoming the limitations of B-mode imaging in distinguishing cancerous from common forms of non-cancerous prostate tissue. Such tissue typing utilizes non-linear methods, such as nearest-neighbor and neural- network techniques, to classify tissues based on spectral- parameter and clinical-variable values. Our research seeks to develop imaging techniques based on these methods for the purpose of improving the guidance of prostate biopsies and the targeting of brachytherapy and external-beam radiotherapy of prostate cancer. Images based on these methods have been imported into real-time instrumentation for biopsy guidance and into commercial dose-planning software for real-time brachytherapy. 3D renderings show locations and volumes of cancer foci. These methods offer exciting possibilities for effective low-cost depiction of prostate cancer in real time and off-line images. Real-time imaging showing cancerous regions of the prostate can be of value in directing biopsies, determining whether biopsy is warranted, assisting in clinical staging, targeting brachytherapy, planning conformal external-beam radiation procedures, and monitoring treatment.

Ultrasonic tissue-type imaging (TTI) for planning treatment of prostate cancer

Our research is intended to develop ultrasonic methods for characterizing cancerous prostate tissue and thereby to improve the effectiveness of biopsy guidance, therapy targeting, and treatment monitoring. We acquired radio-frequency (RF) echo-signal data and clinical variables, e.g., PSA, during biopsy examinations. We computed spectra of the RF signals in each biopsied region, and trained neural network classifers with over 3,000 sets of data using biopsy data as the gold standard. For imaging, a lookup table returned scores for cancer likelihood on a pixel-by-pixel basis from spectral-parameter and PSA values. Using ROC analyses, we compared classification performance of artificial neural networks (ANNs) to conventional classification with a leave-one-patient-out approach intended to minimize the chance of bias. Tissue-type images (TTIs) were compared to prostatectomy histology to further assess classification performance. ROC-curve areas were greater for ANNs than for the B-mode-based classification by more than 20%, e.g., 0.75 +/- 0.03 for neural-networks vs. 0.64 +/- 0.03 for B-mode LOSs. ANN sensitivity was 17% better than the sensitivity range of ultrasound-guided biopsies. TTIs showed tumors that were entirely unrecognized in conventional images and undetected during surgery. We are investigating TTIs for guiding prostrate biopsies, and for planning radiation dose-escalation and tissue-sparing options, and monitoring prostrate cancer.

Advanced ultrasonic tissue-typing and imaging based on radio-frequency spectrum analysis and neural-network classification for guidance of therapy and biopsy procedures

Abstract Conventional B-mode ultrasound is the standard means of imaging the prostate for guiding prostate biopsies and planning brachytherapy of prostate cancer (CaP). Yet B-mode images do not allow visualization of cancerous lesions of the prostate. Ultrasonic tissue-typing imaging based on spectrum analysis of radio-frequency (RF) echo signals has shown promise for overcoming the limitations of B-mode imaging for visualizing prostate tumors. Such tissue-typing utilizes non-linear methods, such as neural-networks, to classify tissue based on spectral-parameter and clinical-variable values. Tissue-type images based on these methods are intended to improve guidance of prostate biopsies and targeting of radiotherapy of prostate cancer. Two-dimensional images have been imported into instrumentation for real-time biopsy guidance and into commercial dose-planning software for brachytherapy planning. Three-dimensional renderings show locations and volumes of cancer foci.

Advances in tissue-type imaging (TTI) for detecting and evaluating prostate cancer

We are seeking to apply advanced ultrasonic tissue typing methods to overcoming severe limitations that now exist in conventional means of imaging cancerous lesions of the prostate. New tissue-type imaging (TTI) methods based on spectrum analysis of radiofrequency (RF) echo signals and neural-network classifiers show potential for distinguishing cancerous from non-cancerous prostate tissues. Encouraging ROC-curve results indicate superior classification. Consequently, these new TTI images may improve biopsy guidance and therapy targeting so that cancers are more-effectively detected and treated.

Advanced ultrasonic imaging of the prostate for guiding biopsies and for planning and monitoring therapy

Our studies are driven by the hypotheses that tissues of the prostate can be classified effectively based on spectrum-analysis parameters computed from ultrasonic radio-frequency echo signals in combination with clinical variables that are evaluated by neural networks and other non-linear methods. The classification of primary interest in these studies is cancerous vs. non-cancerous prostate tissues. Images based on these classification methods would be of clinical value for improving biopsy guidance and non-invasive evaluation of cancer for disease staging, treatment planning, and treatment monitoring. Our classifiers are trained using biopsy results as the gold standard and spectrum-analysis parameters and clinical variables (e.g., PSA value and patient age) as the inputs for tissue typing. ROC curves based on 1,019 biopsies consistently show ROC-curve areas of 0.80 or more for the methods under study compared to areas of 0.66 or less for levels of suspicion based on the B-mode appearance of matching biopsy scan planes. These curves imply that sensitivity improvements of more than 50% might be possible by incorporating these methods into prostate-image encoding. Initial clinical validation of biopsy-guidance efficacy have begun using tissue-classifying images generated in real time; preliminary results suggest that such images increase the likelihood of detecting cancers that are missed when biopsies are guided by conventional ultrasonic images. Parallel studies to incorporate these images into brachytherapy and IMRT planning procedures have been initiated.