Search

BACKGROUND Diffusion-weighted imaging (DWI) is commonly used to detect prostate cancer, and a major clinical challenge is differentiating aggressive from indolent disease. PURPOSE To compare 14 site-specific parametric fitting implementations applied to the same dataset of whole-mount pathologically validated DWI to test the hypothesis that cancer differentiation varies with different fitting algorithms. STUDY TYPE Prospective. POPULATION Thirty-three patients prospectively imaged prior to prostatectomy. FIELD STRENGTH/SEQUENCE 3 T, field-of-view optimized and constrained undistorted single-shot DWI sequence. ASSESSMENT Datasets, including a noise-free digital reference object (DRO), were distributed to the 14 teams, where locally implemented DWI parameter maps were calculated, including mono-exponential apparent diffusion coefficient (MEADC), kurtosis (K), diffusion kurtosis (DK), bi-exponential diffusion (BID), pseudo-diffusion (BID*), and perfusion fraction (F). The resulting parametric maps were centrally analyzed, where differentiation of benign from cancerous tissue was compared between DWI parameters and the fitting algorithms with a receiver operating characteristic area under the curve (ROC AUC). STATISTICAL TEST Levene's test, P

Current MRI signatures of brain cancer often fail to identify regions of hypercellularity beyond the contrast enhancing region. Therefore, this study used autopsy tissue samples aligned to clinical MRIs in order to quantify the relationship between intensity values and cellularity, as well as to develop a radio-pathomic model to predict cellularity using MRI data. This study used 93 samples collected at autopsy from 44 brain cancer patients. Tissue samples were processed, stained for hematoxylin and eosin (HE) and digitized for nuclei segmentation and cell density calculation. Pre- and post-gadolinium contrast T1-weighted images (T1, T1C), T2 fluid-attenuated inversion recovery (FLAIR) images, and apparent diffusion coefficient (ADC) images calculated from diffusion imaging were collected from each patients’ final acquisition prior to death. In-house software was used to align tissue samples to the FLAIR image via manually defined control points. Mixed effect models were used to assess the relationship between single image intensity and cellularity for each image. An ensemble learner was trained to predict cellularity using 5 by 5 voxel tiles from each image, employing a 2/3-1/3 train-test split for validation. Single image analyses found subtle associations between image intensity and cellularity, with a less pronounced relationship within GBM patients. The radio-pathomic model was able to accurately predict cellularity in the test set (RMSE = 1015 cells/mm2) and identified regions of hypercellularity beyond the contrast enhancing region. We concluded that a radio-pathomic model for cellularity is able to identify regions of hypercellular tumor beyond traditional imaging signatures.

Purpose: Multiplex staining allows molecular co-localization analysis of same cell or tissue sections and maximizes the amount of data acquired from individual samples. We developed a non-linear registration framework for automated alignment of whole mount multiplexed histology images of prostate cancer. To achieve a precise automatic fit of bleached and restained high resolution tissue samples, a patch-based approach is proposed to improve annotation speed and analysis. Methods: The three-step co-registration process begins with a coarse low-resolution registration of the IHC stained image to the fixed H&E-stained image. The initial registration is then refined separately for each high-resolution patch using a smaller search window. Finally, registered patches are stitched back together using speeded up robust features (SURF). We apply the method to five multiplex whole mount prostate histology slides. To determine its effectiveness, we compare the automatic registration to the initial coarse registration and a manual control point based-method varying the number of control points. Results: For the control point-based approach, 25, 50, and 100 manually placed set landmarks resulted in a decrease of - 76.20%, -75.9% and -75.48% of the root mean squared error (RMSE), respectively. Compared to the initial registration an improvement of -76.29% of RMSE can be seen, illustrating the potential benefits of a patch-based automatic approach. Conclusion: The proposed method achieved excellent registration of the IHC to the input image. The automated method for registration of multiplexed histology images achieves high accuracy in shorter time and with greater reproducibility than conventional registration approaches and semi-automatic control points without the need for time consuming and subjective manual control-point setting. The accuracy of the fit especially improves in complex areas close to tissue tears and folds.

PURPOSE Tumor treatment fields (TTFields) are approved by the FDA for newly diagnosed as well as recurrent glioblastoma (GBM). TTFields have been shown to extend survival by 4.9 months in newly diagnosed GBM, and a landmark overall survival rate of 13% at 5 years. However, the specific effects remain widely unknown, which has prevented widespread clinical use of this treatment. METHODS This case study examines two glioblastoma patients, IDH-1 wildtype, MGMT unmethylated, that received TTFields (Optune) in addition to maintenance temozolomide (TMZ) following radiation (RT). Both cases were followed using standard MR imaging. Second resections were performed due to radiographic progression of contrast enhancement. RESULTS Although imaging was concerning for tumor progression, pathology showed only treatment effect, ultimately leading to the diagnosis of pseudoprogression. Both patients fell outside the normal expected timeline for chemo-radiation induced pseudoprogression. Based on the pathology, both patients resumed treatment with TMZ/TTFields. One patient expired at 25 months and one is still alive. CONCLUSIONS Pathologic confirmation was essential for guiding further treatment and allowed patients to continue treatment that was effective despite contrary indications on imaging. These findings suggest that pathological confirmation should be strongly considered in patients receiving TMZ/TTFields who develop radiographic progression, potentially with a less invasive biopsy procedure. Future studies should look to characterize the underlying mechanism of interaction between TTFields/TMZ and quantify the prevalence, associated risk factors and whether there is a genotype more susceptible. Both patients reported here had O(6)-methylguanine-DNA methyltransferase (MGMT) unmethylated GBM, and while about 60% of glioblastomas are diagnosed likewise, it is possible that MGMT methylation status plays a role in TTFields response. Better characterization of this phenomenon will improve treatment guidance, potentially reducing unnecessary surgeries and the discontinuance of successful therapies.

This study used large format autopsy tissue samples to compare radio-pathomic maps of brain cancer to a current tumor segmentation algorithm. We hypothesized that an MRI-based machine learning model trained with actual histology rather than radiologist annotations cellularity would 1) improve delineation between tumor and treatment effect, and 2) detect abnormal pathology beyond the contrast-enhancing tumor region. Seventeen patients with pathologically confirmed glioma were included in this study. At autopsy, 43 tissue samples were collected from 17 subjects from whole brain slices sectioned to align with the last axial MRI prior to death. Cellularity was calculated using a color deconvolution on 40X digitized H&E stained slides from the tissue samples. In-house custom software was used to align tissue samples and cellularity information to the FLAIR image using manually defined control points. The DeepMedic algorithm was trained to segment tumors using the BraTs 2017 dataset, and then applied to our patients in order to create automated tumor probability maps. An MRI-based ensemble algorithm using a 5x5 voxel searchlight (input: T1, T1C, FLAIR, ADC) was used to predict cellularity at each voxel, using tissue samples from 14 subjects as ground truth. Both models were applied to 3 withheld test subjects in order to compare tumor probability and cellularity predictions to the pathological ground truth. The mutual information between tumor probability and actual cellularity was 1X10-15, relatively low compared to the rad-path predicted cellularity (=0.16), despite the tumor prediction model accurately highlighting regions of contrast enhancement. Additionally, the radio-pathomic ensemble model correctly identify regions of hypercellularity beyond the tumor segmentation model as well as regions of within the segmented tumor area. This study demonstrates the utility of training machine learning models with pathological ground truth rather than radiologist annotations for predicting localized tumor information, particularly in the post-treatment stage.

Purpose: Prostate cancer primarily arises from the glandular epithelium. Histomophometric techniques have been used to assess the glandular epithelium in automated detection and classification pipelines; however, they are often rigid in their implementation, and their performance suffers on large datasets where variation in staining, imaging, and preparation is difficult to control. The purpose of this study is to quantify performance of a pixelwise segmentation algorithm that was trained using different combinations of weak and strong stroma, epithelium, and lumen labels in a prostate histology dataset. Approach: We have combined weakly labeled datasets generated using simple morphometric techniques and high-quality labeled datasets from human observers in prostate biopsy cores to train a convolutional neural network for use in whole mount prostate labeling pipelines. With trained networks, we characterize pixelwise segmentation of stromal, epithelium, and lumen (SEL) regions on both biopsy core and whole-mount H&E-stained tissue. Results: We provide evidence that by simply training a deep learning algorithm on weakly labeled data generated from rigid morphometric methods, we can improve the robustness of classification over the morphometric methods used to train the classifier. Conclusions: We show that not only does our approach of combining weak and strong labels for training the CNN improve qualitative SEL labeling within tissue but also the deep learning generated labels are superior for cancer classification in a higher-order algorithm over the morphometrically derived labels it was trained on.

Prostate segmentation is a necessary pre-processing step for computer-aided detection and diagnosis algorithms for prostate disorders and associated cancers. Deep learning models like U-Net offer the potential for performing classification and segmentation in a single step. We evaluated the U-Net model for prostate segmentation on 1,235 Magnetic Resonance (MR) images from 39 patients with prostate cancer. On our data set, the U-Net models generated a substantial number of false-positive predictions (average precision of 67.6%). We propose separating classification and segmentation into distinct tasks implemented via a pipeline. Our classifier achieved an average ROC AUC of 97.7% (SD 1.0%) in four-fold cross-fold validation. Compared to the U-Net model alone, our pipeline increased overall agreement between the predicted and human-annotated masks (Dice score of 0.907 vs 0.758 and precision of 83.2% vs 67.6%) without a significant loss in recall (83.4% vs 85.6%). Performing classification and segmentation in a single step is a desirable goal. Our results suggest, however, that improvements are still needed before U-Net offers comparable precision to using a separate classifier.

Prostate cancer is the most common noncutaneous cancer in men in the United States. The current paradigm for screening and diagnosis is imperfect, with relatively low specificity, high cost, and high morbidity. This study aims to generate new image contrasts by learning a distribution of unique image signatures associated with prostate cancer. In total, 48 patients were prospectively recruited for this institutional review board–approved study. Patients underwent multiparametric magnetic resonance imaging 2 weeks before surgery. Postsurgical tissues were annotated by a pathologist and aligned to the in vivo imaging. Radiomic profiles were generated by linearly combining 4 image contrasts (T2, apparent diffusion coefficient [ADC] 0-1000, ADC 50-2000, and dynamic contrast-enhanced) segmented using global thresholds. The distribution of radiomic profiles in high-grade cancer, low-grade cancer, and normal tissues was recorded, and the generated probability values were applied to a naive test set. The resulting Gleason probability maps were stable regardless of training cohort, functioned independent of prostate zone, and outperformed conventional clinical imaging (area under the curve [AUC] = 0.79). Extensive overlap was seen in the most common image signatures associated with high- and low-grade cancer, indicating that low- and high-grade tumors present similarly on conventional imaging.

Multiparametric magnetic resonance imaging (MP-MRI), including diffusion-weighted imaging, is commonly used to diagnose prostate cancer. This radiology–pathology study correlates prostate cancer grade and morphology with common b-value combinations for calculating apparent diffusion coefficient (ADC). Thirty-nine patients undergoing radical prostatectomy were recruited for MP-MRI prior to surgery. Diffusion imaging was collected with seven b-values, and ADC was calculated. Excised prostates were sliced in the same orientation as the MRI using 3-D printed slicing jigs. Whole-mount slides were digitized and annotated by a pathologist. Annotated samples were aligned to the MRI, and ADC values were extracted from annotated peripheral zone (PZ) regions. A receiver operating characteristic (ROC) analysis was performed to determine accuracy of tissue type discrimination and optimal ADC b-value combination. ADC significantly discriminates Gleason (G) G4-5 cancer from G3 and other prostate tissue types. The optimal b-values for discriminating high from low-grade and noncancerous tissue in the PZ are 50 and 2000, followed closely by 100 to 2000 and 0 to 2000. Optimal ADC cut-offs are presented for dichotomized discrimination of tissue types according to each b-value combination. Selection of b-values affects the sensitivity and specificity of ADC for discrimination of prostate cancer.

OBJECTIVES/SPECIFIC AIMS: The objective of this study is to use machine Learning techniques to generate maps of epithelium and lumen density in MRI space. METHODS/STUDY POPULATION: Methods: We prospectively recruited 39 patients undergoing prostatectomy for this institutional review board (IRB) approved study. Patients underwent MP-MRI before prostatectomy on a 3T field strength MRI scanner (General Electric, Waukesha, WI, USA) using an endorectal coil. MP-MRI included field-of-view optimized and constrained undistorted single shot (FOCUS) diffusion weighted imaging with 10 b-values (b=0, 10, 25, 50, 80, 100, 200, 500, 1000, and 2000), dynamic contrast enhanced imaging, and T2-weighted imaging. T2 weighted images were intensity normalized and apparent diffusion coefficient maps were calculated. The dynamic contrast enhanced data was used to calculate the percent change in signal intensity before and after contrast injection. All images were aligned to the T2 weighted image. Robotic prostatectomy was performed 2 weeks after image acquisition. Prostate samples were sliced using a 3D printed slicing jig matching the slice profile of the T2 weighted image. Whole mount samples at 10 μm thickness were taken, hematoxylin and eosin stained, digitized, and annotated by a board certified pathologist. A total of 210 slides were included in this study. Lumen and epithelium were automatically segmented using a custom algorithm written in MATLAB. The algorithm was validated by comparing manual to automatic segmentation on 18 samples. Slides were aligned with the T2 weighted image using a nonlinear control point warping technique. Lumen and epithelium density and the expert annotation were subsequently transformed into MRI space. Co-registration was validated by applying a known warp to tumor masks noted by the pathologist and control point warping the whole mount slide to match the transform. Overlap was measured using a DICE coefficient. A learning curve was generated to determine the optimal number of patients to train the algorithm on. A PLS algorithm was trained on 150 random permutations of patients incrementing from 1 to 29 patients. Slides were stratified such that all slides from a single patient were in the same cohort. Three cohorts were generated, with tumor burden balanced across all cohort. A PLS algorithm was trained on 2 independent training sets (cohorts 1 and 2) and applied to cohort 3. The input vector consisted of MRI values and the target variable was lumen and epithelium density. The algorithm was trained lesion-wise. Trained PiCT models were applied to the test cohort voxel-wise to generate 2 new image contrasts. Mean lesion values were compared between high grade, low grade, and healthy tissue using an ANOVA. An ROC analysis was performed lesion-wise on the test set. RESULTS/ANTICIPATED RESULTS: Results: The segmentation accuracy validation revealed R=0.99 and R=0.72 (p

Magnetic resonance imaging (MRI) is used to diagnose and monitor brain tumors. Extracting additional information from medical imaging and relating it to a clinical variable of interest is broadly defined as radiomics. Here, multiparametric MRI radiomic profiles (RPs) of de novo glioblastoma (GBM) brain tumors is related with patient prognosis. Clinical imaging from 81 patients with GBM before surgery was analyzed. Four MRI contrasts were aligned, masked by margins defined by gadolinium contrast enhancement and T2/fluid attenuated inversion recovery hyperintensity, and contoured based on image intensity. These segmentations were combined for visualization and quantification by assigning a 4-digit numerical code to each voxel to indicate the segmented RP. Each RP volume was then compared with overall survival. A combined classifier was then generated on the basis of significant RPs and optimized volume thresholds. Five RPs were predictive of overall survival before therapy. Combining the RP classifiers into a single prognostic score predicted patient survival better than each alone (P &lt, 005). Voxels coded with 1 RP associated with poor prognosis were pathologically confirmed to contain hypercellular tumor. This study applies radiomic profiling to de novo patients with GBM to determine imaging signatures associated with poor prognosis at tumor diagnosis. This tool may be useful for planning surgical resection or radiation treatment margins.

BACKGROUND Optune therapy with tumor treatment fields (TTFields) is approved for the treatment of recurrent glioblastoma due to a recent clinical trial that showed better quality of life and comparable overall survival to conventional therapy. In the newly diagnosed setting, the addition of TTFields to standard therapy consisting of surgery, radiation and temozolomide has also been shown to prolong tumor progression and improve overall survival. TTFields are low-intensity, alternating frequency electric fields that have been shown to disrupt cell division and subsequently tumor growth. Apoptosis and cell cycle arrest have been seen in vitro, and shown in mice and rabbit tumor models. Though preclinical studies are ongoing, glioblastoma patients who have undergone TTFields therapy have not yet been assessed at autopsy to determine both the pathological signature of TTFields therapy, and the pattern of failure.

Multiparametric magnetic resonance imaging (MP-MRI), including diffusion-weighted imaging, is commonly used to diagnose prostate cancer. This radiology–pathology study correlates prostate cancer grade and morphology with common b-value combinations for calculating apparent diffusion coefficient (ADC). Thirty-nine patients undergoing radical prostatectomy were recruited for MP-MRI prior to surgery. Diffusion imaging was collected with seven b-values, and ADC was calculated. Excised prostates were sliced in the same orientation as the MRI using 3-D printed slicing jigs. Whole-mount slides were digitized and annotated by a pathologist. Annotated samples were aligned to the MRI, and ADC values were extracted from annotated peripheral zone (PZ) regions. A receiver operating characteristic (ROC) analysis was performed to determine accuracy of tissue type discrimination and optimal ADC b-value combination. ADC significantly discriminates Gleason (G) G4-5 cancer from G3 and other prostate tissue types. The optimal b-values for discriminating high from low-grade and noncancerous tissue in the PZ are 50 and 2000, followed closely by 100 to 2000 and 0 to 2000. Optimal ADC cut-offs are presented for dichotomized discrimination of tissue types according to each b-value combination. Selection of b-values affects the sensitivity and specificity of ADC for discrimination of prostate cancer.