Radiomics extracts and mines large number of medical imaging features quantifying tumor phenotypic characteristics Highly accurate and reliable machine-learning approaches can drive the success of radiomic applications in clinical care In this radiomic study, fourteen feature selection methods and twelve classification methods were examined in terms of their performance and stability for predicting overall survival A total of 440 radiomic features were extracted from pre-treatment computed tomography (CT) images of 464 lung cancer patients To ensure the unbiased evaluation of different machine-learning methods, publicly available implementations along with reported parameter configurations were used Furthermore, we used two independent radiomic cohorts for training (n = 310 patients) and validation (n = 154 patients) We identified that Wilcoxon test based feature selection method WLCX (stability = 084 ± 005, AUC = 065 ± 002) and a classification method random forest RF (RSD = 352%, AUC = 066 ± 003) had highest prognostic performance with high stability against data perturbation Our variability analysis indicated that the choice of classification method is the most dominant source of performance variation (3421% of total variance) Identification of optimal machine-learning methods for radiomic applications is a crucial step towards stable and clinically relevant radiomic biomarkers, providing a non-invasive way of quantifying and monitoring tumor-phenotypic characteristics in clinical practice

/pdf/machine-learning-methods-for-quantitative-radiomic-49cj87bvr4.pdf

Machine Learning methods for Quantitative Radiomic Biomarkers

This study aims at developing a joint FDG-PET and MRI texture-based model for the early evaluation of lung metastasis risk in soft-tissue sarcomas (STSs). We investigate if the creation of new composite textures from the combination of FDG-PET and MR imaging information could better identify aggressive tumours. Towards this goal, a cohort of 51 patients with histologically proven STSs of the extremities was retrospectively evaluated. All patients had pre-treatment FDG-PET and MRI scans comprised of T1-weighted and T2-weighted fat-suppression sequences (T2FS). Nine non-texture features (SUV metrics and shape features) and forty-one texture features were extracted from the tumour region of separate (FDG-PET, T1 and T2FS) and fused (FDG-PET/T1 and FDG-PET/T2FS) scans. Volume fusion of the FDG-PET and MRI scans was implemented using the wavelet transform. The influence of six different extraction parameters on the predictive value of textures was investigated. The incorporation of features into multivariable models was performed using logistic regression. The multivariable modeling strategy involved imbalance-adjusted bootstrap resampling in the following four steps leading to final prediction model construction: (1) feature set reduction; (2) feature selection; (3) prediction performance estimation; and (4) computation of model coefficients. Univariate analysis showed that the isotropic voxel size at which texture features were extracted had the most impact on predictive value. In multivariable analysis, texture features extracted from fused scans significantly outperformed those from separate scans in terms of lung metastases prediction estimates. The best performance was obtained using a combination of four texture features extracted from FDG-PET/T1 and FDG-PET/T2FS scans. This model reached an area under the receiver-operating characteristic curve of 0.984 ± 0.002, a sensitivity of 0.955 ± 0.006, and a specificity of 0.926 ± 0.004 in bootstrapping evaluations. Ultimately, lung metastasis risk assessment at diagnosis of STSs could improve patient outcomes by allowing better treatment adaptation.

https://iopscience.iop.org/article/10.1088/0031-9155/60/14/5471/pdf

A radiomics model from joint FDG-PET and MRI texture features for the prediction of lung metastases in soft-tissue sarcomas of the extremities.

CT-based radiomic signature predicts distant metastasis in lung adenocarcinoma

Due to advances in the acquisition and analysis of medical imaging, it is currently possible to quantify the tumor phenotype. The emerging field of Radiomics addresses this issue by converting medical images into minable data by extracting a large number of quantitative imaging features. One of the main challenges of Radiomics is tumor segmentation. Where manual delineation is time consuming and prone to inter-observer variability, it has been shown that semi-automated approaches are fast and reduce inter-observer variability. In this study, a semiautomatic region growing volumetric segmentation algorithm, implemented in the free and publicly available 3D-Slicer platform, was investigated in terms of its robustness for quantitative imaging feature extraction. Fifty-six 3D-radiomic features, quantifying phenotypic differences based on tumor intensity, shape and texture, were extracted from the computed tomography images of twenty lung cancer patients. These radiomic features were derived from the 3D-tumor volumes defined by three independent observers twice using 3D-Slicer, and compared to manual slice-by-slice delineations of five independent physicians in terms of intra-class correlation coefficient (ICC) and feature range. Radiomic features extracted from 3D-Slicer segmentations had significantly higher reproducibility (ICC = 0.85±0.15, p = 0.0009) compared to the features extracted from the manual segmentations (ICC = 0.77±0.17). Furthermore, we found that features extracted from 3D-Slicer segmentations were more robust, as the range was significantly smaller across observers (p = 3.819e-07), and overlapping with the feature ranges extracted from manual contouring (boundary lower: p = 0.007, higher: p = 5.863e-06). Our results show that 3D-Slicer segmented tumor volumes provide a better alternative to the manual delineation for feature quantification, as they yield more reproducible imaging descriptors. Therefore, 3D-Slicer can be employed for quantitative image feature extraction and image data mining research in large patient cohorts.

http://dash.harvard.edu/bitstream/handle/1/12717517/4098900.pdf?sequence=1

Robust Radiomics feature quantification using semiautomatic volumetric segmentation.

Exploring feature-based approaches in PET images for predicting cancer treatment outcomes

A software environment is described, called the computational environment for radiotherapy research (CERR, pronounced "sir"). CERR partially addresses four broad needs in treatment planning research: (a) it provides a convenient and powerful software environment to develop and prototype treatment planning concepts, (b) it serves as a software integration environment to combine treatment planning software written in multiple languages (MATLAB, FORTRAN, C/C++, JAVA, etc.), together with treatment plan information (computed tomography scans, outlined structures, dose distributions, digital films, etc.), (c) it provides the ability to extract treatment plans from disparate planning systems using the widely available AAPM/RTOG archiving mechanism, and (d) it provides a convenient and powerful tool for sharing and reproducing treatment planning research results. The functional components currently being distributed, including source code, include: (1) an import program which converts the widely available AAPM/RTOG treatment planning format into a MATLAB cell-array data object, facilitating manipulation; (2) viewers which display axial, coronal, and sagittal computed tomography images, structure contours, digital films, and isodose lines or dose colorwash, (3) a suite of contouring tools to edit and/or create anatomical structures, (4) dose-volume and dose-surface histogram calculation and display tools, and (5) various predefined commands. CERR allows the user to retrieve any AAPM/RTOG key word information about the treatment plan archive. The code is relatively self-describing, because it relies on MATLAB structure field name definitions based on the AAPM/RTOG standard. New structure field names can be added dynamically or permanently. New components of arbitrary data type can be stored and accessed without disturbing system operation. CERR has been applied to aid research in dose-volume-outcome modeling, Monte Carlo dose calculation, and treatment planning optimization. In summary, CERR provides a powerful, convenient, and common framework which allows researchers to use common patient data sets, and compare and share research results.

CERR: A computational environment for radiotherapy research

IMRT treatment planning for prostate cancer using prioritized prescription optimization and mean-tail-dose functions.

We present middleware for a Client Development Environment that facilitates live development of client applications for SOAP or CORBA servers. We use JPie, a tightly integrated programming environment for live software construction in Java, as the target platform for our design. JPie provides dynamic classes whose signature and implementation can be modified at run time, with changes taking effect immediately upon existing instances of the class. We extend this model to automate addition, mutation, and deletion of dynamic server methods within dynamic clients. Our implementation simplifies distributed application development by masking technical differences between local and remote method invocations. Moreover, the live development model allows server-side changes to be dynamically integrated into a running client to support simultaneous live development of both the client and server.

/pdf/supporting-live-development-of-soap-and-corba-clients-2r3r3h02og.pdf

Supporting Live Development of SOAP and CORBA Clients

Purpose: Radiobiological outcomes models are important predictors of irradiation induced effects in terms of achieving tumor control or causing damage to surrounding normal tissues. They are also used to rank the quality of treatment plans. Outcomes models may depend on many variables such as dose-volume metrics and clinical factors. In particular, the best outcome model itself may vary depending on patient or treatment characteristics. General non-linear models, such as neural-networks, potentially allow us to capture this natural variation in models. Method and Materials: We studied feed-forward (FFNN) and general regression neural networks (GRNN). As representative data, we used a cohort 166 non-small cell lung cancer patients who received radiotherapy treatment, with endpoints of pneumonitis and esophagitis. Dosimetric variables were extracted using CERR. Results: We used resampling (bootstrap) methods to select optimal parameters for the networks, which include the number of neurons in FFNN's and the ‘width’ (σ) in GRNN's. In modeling pneumonitis, the optimal FFNN had 3 layers and 5 neurons in the hidden layer, with spearman rank correlation 0.49±0.27 in training and 0.11±0.07 in testing. The GRNN with σ=1.25, achieved a training spearman of 0.25±0.08 and testing spearman of 0.20±0.3. In modeling esophagitis, the FFNN had 5 neurons, with a spearman of 0.59±0.09 in training and 0.3±0.21 in testing. GRNN with σ=1.25, achieved a training spearman of 0.38±0.06 and a testing spearman of 0.39±0.12. Conclusion: We evaluated two machine learning algorithms to model outcome in cases of pneumonitis and esophagitis. Our preliminary results indicate that the GRNN was more straightforward to implement and, more importantly, had better generalizability than that of FFNN. Our experience to date indicates that neural networks may perform as well or better than multi-term logistic regression methods.

SU‐FF‐T‐375: Machine Learning Methods for Radiobiological Outcomes Modeling

Purpose: Dose‐volume metrics have often been correlated with outcomes and are often used to evaluate treatment plans. Unfortunately, when used for IMRTtreatment planning, dose‐volume metrics are computationally complex (non‐convex) and can warp DVHs near the constraint dose. We investigate whether the generalized equivalent uniform dose (gEUD) can be made to highly correlate with different parts of the DVH curve by tuning the exponential parameter. If so, gEUD may be a smooth and computationally attractive replacement for dose‐volume metrics in treatment planning and evaluation. Method and Materials: We correlated gEUD with various values of its parameter a and clinically applicable dose‐volume constraints. Three datasets were used: lung, esophagus, and prostate, with 219, 263, and 291 patient plans, respectively. We tested values of a between −10 and 10 by intervals of 0.2 and in some cases tested values as low as −40. The dose‐volume constraints tested include: V10, V20, and V30 for lung, V55 for esophagus, and D95 for prostate PTV and lung PTV. Results: For all cases tested, we found a Spearman correlation between 0.917 and 0.989 (mean correlation 0.956) with negligible (<1×10−6) p‐values. Values of a ranged from 0.4 to 3.2 for volume metrics and −7.8 to −27.2 for lung PTV and prostate PTV dose metrics (respectively). Conclusion: There is a significant and strong correlation between dose‐volume metrics and gEUD for the datasets tested. The practical application of this is that for a particular dose‐volume metric, we can find the value of a (the gEUD parameter) with the highest correlation and use the convex gEUD function in place of the non‐convex dose‐volume constraint in the IMRT optimization, thereby allowing optimization to be faster and more able to efficiently achieve a global optimum. Conflict of Interest: Partially supported by NIH grant R01 CA85181 and a grant from TomoTherapy, Inc.

V Clark

Papers

CERR: A computational environment for radiotherapy research

IMRT treatment planning for prostate cancer using prioritized prescription optimization and mean-tail-dose functions.

Supporting Live Development of SOAP and CORBA Clients

SU‐FF‐T‐375: Machine Learning Methods for Radiobiological Outcomes Modeling

TH‐E‐230A‐01: Can Dose‐Volume Parameters Be Replaced with GEUD in the Treatment Planning Process?