NIMG-56. USING RADIOMIC FEATURES FROM DAILY MAGNETIC RESONANCE IMAGING TO PREDICT RESPONSE TO RADIATION THERAPY IN GLIOBLASTOMA PATIENTS: A PILOT STUDY

A dilemma in treating glioblastoma is that MRI after chemotherapy and radiation therapy (chemoRT) shows areas of presumed tumor growth in up to 50% of patients. These areas can represent true progression (TP), tumor growth with tumors non-responsive to treatment, or pseudoprogression (PP), edema and tumor necrosis with favorable treatment response. On imaging, TP and PP are usually not discernable. Patients in this study undergo six weeks of chemoRT on a combination MRI/RT device, receiving daily MRIs. The goal of this study is to explore the correlation of radiomics features with progression. The tumor lesion and surrounding areas of growth/edema were manually outlined as regions of interest (ROIs) for each daily T2-weighted MRI scan. The ROIs were used to calculate texture features: statistical features based on the gray-level co-occurrence matrix (GLCM), the gray-level zone size matrix (GLZSM), the gray-level run length matrix (GLRLM), and the neighborhood gray-tone difference matrix (NGTDM). Each of these matrix classes describe the probability of spatial relationships of gray levels occurring within the ROI. Daily texture features were averaged per week of treatment for each patient. Patient response was retrospectively defined as no progression (NP), TP, or PP. A Kruskal-Wallis test was performed to identify texture features that correlated most strongly with patient response. Forty texture features were calculated for 12 patients (19 treated, 7 excluded due to no T2 lesion or progression status unknown, 6 NP, 3 TP, 3 PP). There was a trend of more texture features correlating significantly with response in weeks 4-6 of treatment, compared to weeks 1-3. A particular texture feature, GLSZM Small Zone Low Gray-Level Emphasis, showed increasing difference between PP and TP over time, with significant difference during week 6 of treatment (p=0.0495). Future directions include correlating early outcomes with greater numbers of patients and daily multiparametric MRI.

Primary clinical study of radiomics for diagnosing simple bone cyst of the jaw

Objective: To screen the radiomic features of simple bone cysts of the jaws and explore the potential application of radiomics in pre-operative diagnosis of jaw simple bone cysts. Methods: The investigators designed and implemented a case–control study. 19 patients with simple bone cysts who were admitted to the Department of Maxillofacial Surgery, Sun Yat-sen University Affiliated Stomatology Hospital from 2013 to 2019 were included in this study. Their clinical data and cone-beam computed tomography (CBCT) images were examined. The control group consisted of patients with odontogenic keratocyst. CBCT imaging features were analyzed and compared between the patient and control groups. Results: Overall, 10,323 image features were extracted through feature analysis. A subset of 25 radiomic features obtained after feature selection were analyzed further. These 25 features were significantly different between the 2 groups (p < 0.05). The absolute value of correlation coefficient was 0.487–0.775. Gray-level co-occurrence matrix (GLCM) contrast, neighborhood gray tone difference matrix (NGTDM) contrast, and GLCM variance were the features with the highest correlation coefficients. Conclusions: Pre-operative radiomics analysis showed the differences between simple bone cysts and odontogenic keratocysts, can help to diagnose simple bone cysts. Three specific texture features—GLCM contrast, NGTDM contrast, and GLCM variance—may be the characteristic imaging features of simple bone cysts of the jaw.

Impact of Lesion Delineation and Intensity Quantisation on the Stability of Texture Features from Lung Nodules on CT: A Reproducible Study

Computer-assisted analysis of three-dimensional imaging data (radiomics) has received a lot of research attention as a possible means to improve the management of patients with lung cancer. Building robust predictive models for clinical decision making requires the imaging features to be stable enough to changes in the acquisition and extraction settings. Experimenting on 517 lung lesions from a cohort of 207 patients, we assessed the stability of 88 texture features from the following classes: first-order (13 features), Grey-level Co-Occurrence Matrix (24), Grey-level Difference Matrix (14), Grey-level Run-length Matrix (16), Grey-level Size Zone Matrix (16) and Neighbouring Grey-tone Difference Matrix (five). The analysis was based on a public dataset of lung nodules and open-access routines for feature extraction, which makes the study fully reproducible. Our results identified 30 features that had good or excellent stability relative to lesion delineation, 28 to intensity quantisation and 18 to both. We conclude that selecting the right set of imaging features is critical for building clinical predictive models, particularly when changes in lesion delineation and/or intensity quantisation are involved.

Assessment of genetic differentiation of populations by analysis of molecular variance (analytical review)

Different approaches to using the analysis of molecular variance (AMOVA) to assess the genetic differentiation of populations have been compared in the research. Data on 11 microsatellite loci of 84 bulls of seven breeds were used. The results were compared for three options of the AMOVA module of the GenAlEx 6.502 program: the allele distance matrix (calculated FST(W&C)(=θ) statistics – variant AMOVA1); the genotype distance matrix (ΦPT– AMOVA2); and the allele size difference matrix (RST– AMOVA3). Similar summary estimates of the genetic differentiation of breeds were obtained: FST(W&C)= 0.108, ΦPT= 0.115, RST= 0.110 (all with pperm≤ 0.001). Between the estimates of FST(W&C)and ΦPTfor each locus, the correlation coefficient was 0.99 (pvalue<0.0001); no statistically significant correlations with RSTwere found. A high correlation of FST(W&C)and ΦPTwith the estimates of differentiation according to Nei’s (0.96) was found. Programs other than GenAlEx (Arlequin v.3.5, GenePop v.4.7.3, RST22) gave similar AMOVA estimates. The negative linear dependence of FST(W&C)and ΦPTon the level of the average heterozygosity of the breed samples was established (R2= 0.6, rS= -0.75 for pvalue < 0.02) and the absence of such dependence for RST(R2= 0.04, rS= -0.23 for pvalue= 0.47). The standardization of the FST(W&C)and ΦPTestimates according to Hedrick’s eliminated this dependence and raised the initial estimates to 0.35 and 0.37, respectively. The latter were comparable to the estimates obtained by the Nei-Hedrick’s (0.364-0.375), Jost’s (0.292), and Morisit-Horn’s (0.308) methods. The Mantel correlations between the matrices of paired genetic distances (GD) calculated by different measures were >0.9 in most cases. The projections of the GD matrices in the principal coordinate analysis (PCoA) on the 2D plane were generally similar. The PCoA identified a cluster of Holstein «ecotypes», a cluster of «Red» breeds, and a branch of the Jersey breed. In the two-factor AMOVA of data on clusters (as two «regions»), the interregional GD was 0.357; the differentiation of breeds within the «regions» did not exceed 0.027. Modeling the association of breeds with close to zero GD resulted in an increase in the number of alleles per locus in the «new» breeds by 29 %, and an increase in the combined estimate of genetic differentiation by 29-46 %. The results obtained can be used in the development of measures for the conservation of endangered breeds.

Are computed-tomography-based hematoma radiomics features reproducible and predictive of intracerebral hemorrhage expansion? an in vitro experiment and clinical study

Objectives: To identify reproducible hematoma radiomics features (RFs) for use in predicting hematoma expansion (HE) in patients with acute intracerebral hemorrhage (ICH). Methods: For test–retest analysis, three syringes with different volumes of blood collected at the same time (to mimic homogeneous hematoma) and a phantom (FT/HK 2000; Huake, Szechwan, China) containing three cylindrical inserts were scanned seven times within 6 h on the same CT scanner. Three additional syringes with mixed blood collected at different time points (to mimic heterogeneous hematoma) were tied together with the first three syringes as well as the phantom were scanned using modified CT acquisition parameters for intra CT analysis. A coefficient of variation below 10% served as the cutoff value for reproducibility. Finally, reproducible and potentially useful RFs were used to predict HE in 144 acute ICH patients, with the area under the receiver operating characteristic curves (AUC) used to evaluate their diagnostic performance. Results: A total of 630 RFs including 18 first-order, 24 gray-level co-occurrence matrix (GLCM), 16 gray-level run length matrix (GLRLM), five neighborhood gray-tone difference matrix (NGTDM), 63 Laplacian of Gaussian (LoG), and 504 Wavelet features were evaluated. In the test-retest analysis, the percentages of reproducible RFs ranged from 42.54% (268/630) to 45.4% (286/630) for the three homogeneous hematoma samples and 79.05% (498/630) to 81.43% (513/630) for the phantom. In the intra-CT analysis, the percentages varied from 31.43% (198/630) to 42.38% (267/630) for the six hematoma samples and 48.89% (308/630) to 53.97% (340/630) for the phantom. In the in vitro experiment, 148 RFs were reproducible for all hematoma samples in both the test-retest and intra-CT analyses; however, only 80 were statistically different between homogeneous and heterogeneous hematoma samples. Finally, HE occurred in 25% (growth >6 ml, 36/144) to 31.94% (growth >3 ml or 33%, 46/144) of the patients. The AUCs in predicting HE ranged from 0.625 to 0.703. Conclusions: Only a few CT-based RFs from the in vitro hematoma were reproducible and can distinguish between homogeneous and heterogeneous hematomas. The use of RFs alone to predict HE in acute ICH showed only a moderate performance. Advances in knowledge: Using an in vitro experiment and clinical validation, this study demonstrated for the first time that CT-based hematoma RFs can be used to predict HE in acute ICH; nonetheless, only a few RFs are reproducible and can be used for prediction.

A Data-Driven Methodology to Improve Tolerance Allocation Using Product Usage Data

Industry 4.0, the fourth industrial revolution, puts forward new requirements for the sustainable service of products. With the recent advances in measurement technologies, global and local deformations in inaccessible areas can be monitored. Product usage data such as geometric deviation, position deviation, and angular deviation that lead to product functional performance degradation can be continuously collected during the product usage stage. These technologies provide opportunities to improve tolerance design by improving tolerance allocation using product usage data. The challenge lies in how to assess these deviations for identifying relevant field factors and reallocate the tolerance value. In this paper, a data-driven methodology based on the deviation for tolerance analysis is proposed to improve the tolerance allocation. A feature graph of a mechanical assembly is established based on the assembly relationship. The node representation in the feature graph is defined based on the unified Jacobian-torsor model and the node label is calculated by a synthetic evaluation method. A novel hierarchical graph attention networks (HGAT) is proposed to investigate hidden relations between nodes in the feature graph and calculate labels of all nodes. A modification necessity index (MNI) is defined for each tolerance between two nodes based on their labels. An identification of the to-be-modified tolerance method is proposed to specify the tolerance analysis target. A deviation difference matrix is constructed to calculate the MNI of each tolerance for identifying the to-be-modified tolerance value with high priorities for product improvement. The effectiveness of the proposed methodology is demonstrated through a case study for improving tolerance allocation of a press machine.

Use of Texture Analysis on Noncontrast MRI in Classification of Early Stage of Liver Fibrosis

Purpose. To compare the diagnostic value of texture analysis- (TA-) derived parameters from out-of-phase T1W, in-phase T1W, and T2W images in the classification of the early stage of liver fibrosis. Methods. Patients clinically diagnosed with hepatitis B infection, who underwent liver biopsy and noncontrast MRI scans, were enrolled. TA parameters were extracted from out-of-phase T1-weighted (T1W), in-phase T1W, and T2-weighted (T2W) images and calculated using Artificial Intelligent Kit (AK). Features were extracted including first-order, shape, gray-level cooccurrence matrix, gray-level run-length matrix, neighboring gray one tone difference matrix, and gray-level differential matrix. After statistical analyses, final diagnostic models were constructed. Receiver operating curves (ROCs) and areas under the ROC (AUCs) were used to assess the diagnostic value of each final model and 100-time repeated cross-validation was applied to assess the stability of the logistic regression models. Results. A total of 57 patients were enrolled in this study, with 27 in the fibrosis stage < 2 and 30 in stages ≥ 2. Overall, 851 features were extracted per ROI. Eight features with high correlation were selected by the maximum relevance method in each sequence, and all had a good diagnostic performance. ROC analysis of the final models showed that all sequences had a preferable performance with AUCs of 0.87, 0.90, and 0.96 in T2W and in-phase and out-of-phase T1W, respectively. Cross-validation results reported the following values of mean accuracy, specificity, and sensitivity: 0.98 each for out-of-phase T1W; 0.90, 0.89, and 0.90 for in-phase T1W; and 0.86, 0.88, 0.84 for T2W in the training set, and 0.76, 0.81, and 0.72 for out-of-phase T1W; 0.74, 0.72, and 0.75 for in-phase T1W; and 0.63, 0.64, and 0.63 for T2W for the test group, respectively. Conclusion. Noncontrast MRI scans with texture analysis are viable for classifying the early stages of liver fibrosis, exhibiting excellent diagnostic performance.

Comparison of Radiomic Models Based on Low-Dose and Standard-Dose CT for Prediction of Adenocarcinomas and Benign Lesions in Solid Pulmonary Nodules

ObjectivesThis study aimed to develop radiomic models based on low-dose CT (LDCT) and standard-dose CT to distinguish adenocarcinomas from benign lesions in patients with solid solitary pulmonary nodules and compare the performance among these radiomic models and Lung CT Screening Reporting and Data System (Lung-RADS). The reproducibility of radiomic features between LDCT and standard-dose CT were also evaluated.MethodsA total of 141 consecutive pathologically confirmed solid solitary pulmonary nodules were enrolled including 50 adenocarcinomas and 48 benign nodules in primary cohort and 22 adenocarcinomas and 21 benign nodules in validation cohort. LDCT and standard-dose CT scans were conducted using same acquisition parameters and reconstruction method except for radiation dose. All nodules were automatically segmented and 104 original radiomic features were extracted. The concordance correlation coefficient was used to quantify reproducibility of radiomic features between LDCT and standard-dose CT. Radiomic features were selected to build radiomic signature, and clinical characteristics and radiomic signature were combined to develop radiomic nomogram for LDCT and standard-dose CT, respectively. The performance of radiomic models and Lung-RADS was assessed by area under curve (AUC) of receiver operating characteristic curve, sensitivity, and specificity.ResultsShape and first order features, and neighboring gray tone difference matrix features were highly reproducible between LDCT and standard-dose CT. No significant differences of AUCs were found among radiomic signature and nomogram of LDCT and standard-dose CT in both primary and validation cohort (0.915 vs. 0.919 vs. 0.898 vs. 0.909 and 0.976 vs. 0.976 vs. 0.985 vs. 0.987, respectively). These radiomic models had higher specificity than Lung-RADS (all correct P &lt; 0.05), while there were no significant differences of sensitivity between Lung-RADS and radiomic models.ConclusionsThe diagnostic performance of LDCT-based radiomic models to differentiate adenocarcinomas from benign lesions in solid pulmonary nodules were equivalent to that of standard-dose CT. The LDCT-based radiomic model with higher specificity and lower false-positive rate than Lung-RADS might help reduce overdiagnosis and overtreatment of solid pulmonary nodules in lung cancer screening.

BP neural Network-Kalman filter fusion method for unmanned aerial vehicle target tracking

The poor real-time performance and target occlusion occurred easily when the UAV was tracking the target. In this paper, a target tracking method based on the Back Propagation neural network fusion Kalman filter algorithm was developed to solve the position prediction problem of the UAV target tracking in real time. Firstly, the target tracking algorithm was used to acquire the center position coordinates of the target on the onboard computer, and then the coordinate difference matrix was constructed to train the BP neural network in real time. Secondly, when the target was occluded by the obstacles judged by the Bhattacharyya coefficient, the BP neural network fusion Kalman filter algorithm was used to accurately predict the center position coordinates of the occluded target. Then the flight speed of UAV was calculated by the deviation between the coordinates of the target and the image center. Finally, the velocity command was sent to the UAV by the onboard computer. The experimental results shown that the target position predicted by BP neural network fusion Kalman filter algorithm was more accurate and robust in predicting the center position coordinates of the target, and the UAV can track the moving target on the ground stably.