Segmentation of 3-D High-Frequency Ultrasound Images of Human Lymph Nodes Using Graph Cut With Energy Functional Adapted to Local Intensity Distribution.

Previous studies by our group have shown that 3-D high-frequency quantitative ultrasound (QUS) methods have the potential to differentiate metastatic lymph nodes (LNs) from cancer-free LNs dissected from human cancer patients. To successfully perform these methods inside the LN parenchyma (LNP), an automatic segmentation method is highly desired to exclude the surrounding thin layer of fat from QUS processing and accurately correct for ultrasound attenuation. In high-frequency ultrasound images of LNs, the intensity distribution of LNP and fat varies spatially because of acoustic attenuation and focusing effects. Thus, the intensity contrast between two object regions (e.g., LNP and fat) is also spatially varying. In our previous work, nested graph cut (GC) demonstrated its ability to simultaneously segment LNP, fat, and the outer phosphate-buffered saline bath even when some boundaries are lost because of acoustic attenuation and focusing effects. This paper describes a novel approach called GC with locally adaptive energy to further deal with spatially varying distributions of LNP and fat caused by inhomogeneous acoustic attenuation. The proposed method achieved Dice similarity coefficients of 0.937±0.035 when compared with expert manual segmentation on a representative data set consisting of 115 3-D LN images obtained from colorectal cancer patients.

Local Transverse-Slice-Based Level-Set Method for Segmentation of 3-D High-Frequency Ultrasonic Backscatter From Dissected Human Lymph Nodes.

OBJECTIVE: To detect metastases in freshly excised human lymph nodes (LNs) using three-dimensional (3-D), high-frequency, quantitative ultrasound (QUS) methods, the LN parenchyma (LNP) must be segmented to preclude QUS analysis of data in regions outside the LNP and to compensate ultrasound attenuation effects due to overlying layers of LNP and residual perinodal fat (PNF). METHODS: After restoring the saturated radio-frequency signals from PNF using an approach based on smoothing cubic splines, the three regions, i.e., LNP, PNF, and normal saline (NS), in the LN envelope data are segmented using a new, automatic, 3-D, three-phase, statistical transverseslice-based level-set (STS-LS) method that amends Lankton's method. Due to ultrasound attenuation and focusing effects, the speckle statistics of the envelope data vary with imaged depth. Thus, to mitigate depth-related inhomogeneity effects, the STS-LS method employs gamma probabilitydensity functions to locally model the speckle statistics within consecutive transverse slices. RESULTS: Accurate results were obtained on simulated data. On a representative dataset of 54 LNs acquired from colorectal-cancer patients, the Dice similarity coefficient for LNP, PNF, and NS were 0.938 ± 0.025, 0.832 ± 0.086, and 0.968 ± 0.008, respectively, when compared to expert manual segmentation. CONCLUSION: The STS-LS outperforms the established methods based on global and local statistics in our datasets and is capable of accurately handling the depth-dependent effects due to attenuation and focusing. SIGNIFICANCE: This advance permits the automatic QUS-based cancer detection in the LNs. Furthermore, the STS-LS method could potentially be used in a wide range of ultrasound-imaging applications suffering from depth-dependent effects.

Modeling the envelope statistics of three-dimensional high-frequency ultrasound echo signals from dissected human lymph nodes.

This work investigates the statistics of the envelope of three-dimensional (3D) high-frequency ultrasound (HFU) data acquired from dissected human lymph nodes (LNs). Nine distributions were employed, and their parameters were estimated using the method of moments. The Kolmogorov Smirnov (KS) metric was used to quantitatively compare the fit of each candidate distribution to the experimental envelope distribution. The study indicates that the generalized gamma distribution best models the statistics of the envelope data of the three media encountered: LN parenchyma, fat and phosphate-buffered saline (PBS). Furthermore, the envelope statistics of the LN parenchyma satisfy the pre-Rayleigh condition. In terms of high fitting accuracy and computationally efficient parameter estimation, the gamma distribution is the best choice to model the envelope statistics of LN parenchyma, while, the Weibull distribution is the best choice to model the envelope statistics of fat and PBS. These results will contribute to the development of more-accurate and automatic 3D segmentation of LNs for ultrasonic detection of clinically significant LN metastases.

Three-dimensional quantitative ultrasound for detecting lymph node metastases.

PURPOSE: Detection of metastases in lymph nodes (LNs) is critical for cancer management. Conventional histological methods may miss metastatic foci. To date, no practical means of evaluating the entire LN volume exists. The aim of this study was to develop fast, reliable, operator-independent, high-frequency, quantitative ultrasound (QUS) methods for evaluating LNs over their entire volume to effectively detect LN metastases. METHODS: We scanned freshly excised LNs at 26 MHz and digitally acquired echo-signal data over the entire three-dimensional (3D) volume. A total of 146 LNs of colorectal, 26 LNs of gastric, and 118 LNs of breast cancer patients were enrolled. We step-sectioned LNs at 50-µm intervals and later compared them with 13 QUS estimates associated with tissue microstructure. Linear-discriminant analysis classified LNs as metastatic or nonmetastatic, and we computed areas (Az) under receiver-operator characteristic curves to assess classification performance. The QUS estimates and cancer probability values derived from discriminant analysis were depicted in 3D images for comparison with 3D histology. RESULTS: Of 146 LNs of colorectal cancer patients, 23 were metastatic; Az = 0.952 ± 0.021 (95% confidence interval [CI]: 0.911-0.993); sensitivity = 91.3% (specificity = 87.0%); and sensitivity = 100% (specificity = 67.5%). Of 26 LNs of gastric cancer patients, five were metastatic; Az = 0.962 ± 0.039 (95% CI: 0.807-1.000); sensitivity = 100% (specificity = 95.3%). A total of 17 of 118 LNs of breast cancer patients were metastatic; Az = 0.833 ± 0.047 (95% CI: 0.741-0.926); sensitivity = 88.2% (specificity = 62.5%); sensitivity = 100% (specificity = 50.5%). 3D cancer probability images showed good correlation with 3D histology. CONCLUSIONS: These results suggest that operator- and system-independent QUS methods allow reliable entire-volume LN evaluation for detecting metastases. 3D cancer probability images can help pathologists identify metastatic foci that could be missed using conventional methods.

Three-dimensional quantitative ultrasound to guide pathologists towards metastatic foci in lymph nodes.

The detection of metastases in freshly-excised lymph nodes from cancer patients during lymphadenectomy is critically important for cancer staging, treatment, and optimal patient management. Currently, conventional histologic methods suffer a high rate of false-negative determinations because pathologists cannot evaluate each excised lymph nodes in its entirety. Therefore, lymph nodes are undersampled and and small but clinically relevant metastatic regions can be missed. In this study, quantitative ultrasound (QUS) methods using high-frequency transducers (i.e., > 20 MHz) were developed and evaluated for their ability to detect and guide pathologists towards suspicious regions in lymph nodes. A custom laboratory scanning system was used to acquire radio-frequency (RF) data in 3D from excised lymph nodes using a 26-MHz center-frequency transducer. Overlapping 1-mm cylindrical regions-of-interest (ROIs) of the RF data were processed to yield 13 QUS estimates quantifying tissue microstructure and organization. These QUS methods were applied to more than 260 nodes from more than 160 colorectal-, gastric-, and breast-cancer patients. Cancer-detection performance was assessed for individual estimates and linear combinations of estimates. ROC results demonstrated excellent classification. For colorectal- and gastric-cancer nodes, the areas under the ROC curves (AUCs) were greater than 0.95. Slightly poorer results (AUC=0.85) were obtained for breast-cancer nodes. Images based on QUS parameters also permitted localization of cancer foci in some micrometastatic cases.

Entire-volume serial histological examination for detection of micrometastases in lymph nodes of colorectal cancers.

The purpose of this study was to accurately detect lymph-node micrometastases, i.e., metastatic cancer foci that have a size between 2.0 and 0.2 mm, in nodes excised from colorectal cancer (CRC) patients, and to determine how frequently micrometastases might be missed when standard histological examination procedures are used. A total of 311 lymph nodes were removed and examined from 90 patients with Stage I to IV CRC. The number of slices of histology sections ranged from 6 to 75 per node (average = 25.5; SD = 11.1), which provided a total of 7,943 slices. Lymph nodes were examined in their entire volume at every 50-µm and 100-µm intervals for nodes smaller and larger than 5 mm respectively. The total number of thin sections examined in each node and the number of thin sections where metastatic foci were present were counted. The number of thin sections with metastatic foci and the total number of slices was determined for each node. In addition, the presence or absence of metastatic foci in the "central" slice was determined. Micrometastases were found in 12/311 (3.9%) of all lymph nodes. In the 12 lymph nodes with micrometastases, the rate of metastatic slices over all slices was 39.4% (range = 6.3 to 81.3%; SD = 25.8%) In the central slice of each node, micrometastases were present only in 6 of 12 lymph nodes (50%); accordingly, they were not present in the central slice for half the micrometastatic nodes. These 6 nodes represented 1.9% of the 311 nodes and 11.1% of the 54 metastatic nodes. This study suggests that a significant fraction of micrometastases can be missed by traditional singleslice sectioning; half of the micrometastases would have been overlooked in our data set of 311 nodes.

Three-dimensional high-frequency backscatter and envelope quantification of cancerous human lymph nodes.

Quantitative imaging methods using high-frequency ultrasound (HFU) offer a means of characterizing biological tissue at the microscopic level. Previously, high-frequency, 3-D quantitative ultrasound (QUS) methods were developed to characterize 46 freshly-dissected lymph nodes of colorectal-cancer patients. 3-D ultrasound radiofrequency data were acquired using a 25.6 MHz center-frequency transducer and each node was inked before tissue fixation to recover orientation after sectioning for 3-D histological evaluation. Backscattered echo signals were processed using 3-D cylindrical regions-of-interest (ROIs) to yield four QUS estimates associated with tissue microstructure (i.e., effective scatterer size, acoustic concentration, intercept and slope). These QUS estimates, obtained by parameterizing the backscatter spectrum, showed great potential for cancer detection. In the present study, these QUS methods were applied to 112 lymph nodes from 77 colorectal and gastric cancer patients. Novel QUS methods parameterizing the envelope statistics of the ROIs using Nakagami and homodyned-K distributions were also developed; they yielded four additional QUS estimates. The ability of these eight QUS estimates to classify lymph nodes and detect cancer was evaluated using receiver operating characteristics (ROC) curves. An area under the ROC curve of 0.996 with specificity and sensitivity of 95% were obtained by combining effective scatterer size and one envelope parameter based on the homodyned-K distribution. Therefore, these advanced 3-D QUS methods potentially can be valuable for detecting small metastatic foci in dissected lymph nodes.