Extracting Grain Orientations from EBSD Patterns of Polycrystalline Materials Using Convolutional Neural Networks.

We present a deep learning approach to the indexing of electron backscatter diffraction (EBSD) patterns. We design and implement a deep convolutional neural network architecture to predict crystal orientation from the EBSD patterns. We design a differentiable approximation to the disorientation function between the predicted crystal orientation and the ground truth; the deep learning model optimizes for the mean disorientation error between the predicted crystal orientation and the ground truth using stochastic gradient descent. The deep learning model is trained using 374,852 EBSD patterns of polycrystalline nickel from simulation and evaluated using 1,000 experimental EBSD patterns of polycrystalline nickel. The deep learning model results in a mean disorientation error of 0.548° compared to 0.652° using dictionary based indexing.

Moving closer to experimental level materials property prediction using AI.

While experiments and DFT-computations have been the primary means for understanding the chemical and physical properties of crystalline materials, experiments are expensive and DFT-computations are time-consuming and have significant discrepancies against experiments. Currently, predictive modeling based on DFT-computations have provided a rapid screening method for materials candidates for further DFT-computations and experiments; however, such models inherit the large discrepancies from the DFT-based training data. Here, we demonstrate how AI can be leveraged together with DFT to compute materials properties more accurately than DFT itself by focusing on the critical materials science task of predicting "formation energy of a material given its structure and composition". On an experimental hold-out test set containing 137 entries, AI can predict formation energy from materials structure and composition with a mean absolute error (MAE) of 0.064 eV/atom; comparing this against DFT-computations, we find that AI can significantly outperform DFT computations for the same task (discrepancies of [Formula: see text] eV/atom) for the first time.

Generative Adversarial Networks and Mixture Density Networks-Based Inverse Modeling for Microstructural Materials Design.

There are two broad modeling paradigms in scientific applications: forward and inverse. While forward modeling estimates the observations based on known causes, inverse modeling attempts to infer the causes given the observations. Inverse problems are usually more critical as well as difficult in scientific applications as they seek to explore the causes that cannot be directly observed. Inverse problems are used extensively in various scientific fields, such as geophysics, health care and materials science. Exploring the relationships from properties to microstructures is one of the inverse problems in material science. It is challenging to solve the microstructure discovery inverse problem, because it usually needs to learn a one-to-many nonlinear mapping. Given a target property, there are multiple different microstructures that exhibit the target property, and their discovery also requires significant computing time. Further, microstructure discovery becomes even more difficult because the dimension of properties (input) is much lower than that of microstructures (output). In this work, we propose a framework consisting of generative adversarial networks and mixture density networks for inverse modeling of structure-property linkages in materials, i.e., microstructure discovery for a given property. The results demonstrate that compared to baseline methods, the proposed framework can overcome the above-mentioned challenges and discover multiple promising solutions in an efficient manner.

Enabling deeper learning on big data for materials informatics applications.

The application of machine learning (ML) techniques in materials science has attracted significant attention in recent years, due to their impressive ability to efficiently extract data-driven linkages from various input materials representations to their output properties. While the application of traditional ML techniques has become quite ubiquitous, there have been limited applications of more advanced deep learning (DL) techniques, primarily because big materials datasets are relatively rare. Given the demonstrated potential and advantages of DL and the increasing availability of big materials datasets, it is attractive to go for deeper neural networks in a bid to boost model performance, but in reality, it leads to performance degradation due to the vanishing gradient problem. In this paper, we address the question of how to enable deeper learning for cases where big materials data is available. Here, we present a general deep learning framework based on Individual Residual learning (IRNet) composed of very deep neural networks that can work with any vector-based materials representation as input to build accurate property prediction models. We find that the proposed IRNet models can not only successfully alleviate the vanishing gradient problem and enable deeper learning, but also lead to significantly (up to 47%) better model accuracy as compared to plain deep neural networks and traditional ML techniques for a given input materials representation in the presence of big data.

Author Correction: Enhancing materials property prediction by leveraging computational and experimental data using deep transfer learning.

An amendment to this paper has been published and can be accessed via a link at the top of the paper.

Enhancing materials property prediction by leveraging computational and experimental data using deep transfer learning.

The current predictive modeling techniques applied to Density Functional Theory (DFT) computations have helped accelerate the process of materials discovery by providing significantly faster methods to scan materials candidates, thereby reducing the search space for future DFT computations and experiments. However, in addition to prediction error against DFT-computed properties, such predictive models also inherit the DFT-computation discrepancies against experimentally measured properties. To address this challenge, we demonstrate that using deep transfer learning, existing large DFT-computational data sets (such as the Open Quantum Materials Database (OQMD)) can be leveraged together with other smaller DFT-computed data sets as well as available experimental observations to build robust prediction models. We build a highly accurate model for predicting formation energy of materials from their compositions; using an experimental data set of [Formula: see text] observations, the proposed approach yields a mean absolute error (MAE) of [Formula: see text] eV/atom, which is significantly better than existing machine learning (ML) prediction modeling based on DFT computations and is comparable to the MAE of DFT-computation itself.

ElemNet: Deep Learning the Chemistry of Materials From Only Elemental Composition.

Conventional machine learning approaches for predicting material properties from elemental compositions have emphasized the importance of leveraging domain knowledge when designing model inputs. Here, we demonstrate that by using a deep learning approach, we can bypass such manual feature engineering requiring domain knowledge and achieve much better results, even with only a few thousand training samples. We present the design and implementation of a deep neural network model referred to as ElemNet; it automatically captures the physical and chemical interactions and similarities between different elements using artificial intelligence which allows it to predict the materials properties with better accuracy and speed. The speed and best-in-class accuracy of ElemNet enable us to perform a fast and robust screening for new material candidates in a huge combinatorial space; where we predict hundreds of thousands of chemical systems that could contain yet-undiscovered compounds.

Significance of CA19-9 in predicting the prognosis of urothelial carcinoma: a hospital based study from Nepal.

BACKGROUND: The present study was undertaken to establish any correlation of elevated levels of CA19-9 with tumor stage or grade of urothelial carcinoma. MATERIALS AND METHODS: This hospital based study was carried out in the Department of Biochemistry of Nepalese Army Institute of Health Sciences between 1st July 2012 and 31st December 2012. Approval for the study was obtained from the institutional research ethical committee. CA19-9 was assayed with an ELISA reader for all cases and expressed in U/ml with 37U/ml taken as the cut-off upper value for normal. RESULTS: Out of 20 cases enrolled, 15 were of urothelial carcinoma and the remaining 5 were controls. There was marked difference between the mean values of CA19-9 in cases 40.2±19.3U/ml of urothelial carcinoma and controls 7.98±7.34U/ml. The number of cases in Ta, TI, T2, T3, T4 stages of urothelial carcinoma were 2, 6, 3, 3, 1 respectively. The percentage rise in CA19-9 was less with low grade tumors (22.2%) when compared with high grade tumors (66.6%) (p value 0.001*). The percentage of rise in CA19-9 for muscle invasive tumors was very high when compared to superficial tumors. Similarly, the percentage of rise in CA19- 9 for metastatic disease was very high when compared to non-metastatic disease and it was found statistically significant (p value 0.001*). CONCLUSION: Serum CA19-9 levels predicts the prognosis of urothelial carcinoma as it is almost invariably raised in tumors having metastatic spread.

Elevated serum insulin is an independent risk factor for hepatocellular carcinoma: a case control study from Nepal.

AIM: To investigate associations of fasting insulin and glucose levels in serum with hepatocellular carcinoma risk. MATERIALS AND METHODS: This hospital based study was carried out using data retrieved from the register maintained in the Department of Biochemistry of the Nepalese Army Institute of Health Sciences, between 1st December, 2011 and 31st June, 2013. The variables collected were age, fasting plasma glucose, fasting plasma insulin and ALT. Quantitative determination of human insulin concentrations was accomplished by chemiluminescence enzyme immunoassay. RESULTS: Of the total 220 subjects enrolled in our present study, 20 cases were of HCC and 200 were healthy controls. The maximum number of cases of hepatocellular carcinoma in category cutpoints of fasting insulin levels fell in the range of >6.10 µU/ml. The highest insulin levels (>6.10 µU/ml) were seen to be associated with an 2.36 fold risk of HCC when compared with fasting insulin levels of (<2.75 µU/ml). Furthermore, the insulin levels (2.75-4.10 µU/ml) of category cutpoints also conferred a 1.57 fold risk for HCC when compared with lowest fasting insulin levels of (<2.75 µU/ml). CONCLUSIONS: The effect of an insulin level in increasing HCC risk appeared consistent, influencing incidence, risk of recurrence, overall survival, and treatment-related complications in HCC patients.

Impact of various tumor markers in prognosis of gastric cancer. A hospital based study from tertiary care hospital of Kathmandu valley.

BACKGROUND: To obtain the maximum additional information about the prognosis of gastric cancer, we compared CA-50 with other previously defined markers. MATERIALS AND METHODS: This hospital based study was carried out in the Department of Biochemistry of Nepalese Army Institute of Health Sciences between 1st July 2012 and 31st December 2012. The variables collected were age, gender, AFP, CEA, CA19-9, and CA50, assayed with ELISA reader for all cases. The cut off values for serum AFP, CEA, CA19-9, and CA-50 were 10 µg/l, 10 µg/l, 37 U/ml, and 20 U/ml, respectively according to the manufacturer's instructions. Approval for the study was obtained from the institutional research ethical committee. RESULTS: Of the 40 examined patients, 13 patients had tumors located in the upper third of the stomach, 6 patients had tumors in the middle third, 16 patients had tumors in the lower third, and 5 patients had tumors occupying two-thirds of the stomach or more. The distribution of lymph node staging of the patients was as follows: 7 patients belonged to N0, 9 patients to N1 stage, 10 patients to N2 stage, and 14 patients to N3 stage. The statistical method of Cox proportional hazards using multivariate analysis also illustrated that tumor markers including CEA (2.802), CA19-9 (2.690), CA50 (2.101), were independent prognostic factors, as tumor size (1.603), and lymph node stage (1.614). CONCLUSIONS: The tumour markers now available, like CEA, CA 19-9 and CA 50, chiefly perceive advanced gastric cancer. The preoperative rise in those tumour marker level have a prognostic significance and may be clinically helpful in choosing patients for adjuvant management.

Diabetes mellitus and renal cell carcinoma--a hospital based study from Kathmandu Valley.

OBJECTIVE: To diagnose renal cell carcinoma at early stages and for better prognosis , the main objective of our current study was to understand any association with diabetes with relation to age, gender, history of disease, diabetic laboratory parameters, tumor size and grade. MATERIALS AND METHODS: This hospital based study was carried out using data retrieved from the register maintained in the Department of Biochemistry of Nepalese Army Institute of Health Sciences between 1st December, 2011 and 31st May, 2012. The variables collected were age, gender, HbA1c, serum creatinine, fasting blood glucose. One way ANOVA was applied to examine statistical significance of differences between groups. The LSD post hoc test was used for the comparison of means of case groups. RESULTS: Of the total 140 cases of renal cell carcinoma, 79 patients were also suffering from diabetes mellitus. The number of females (47) was more in diabetic RCC patients when compared to males (32). Significance was observed in levels of serum creatinine for tumor size >10 cm (0.0001*). The highest value of glycated hemoglobin (8.9%) and fasting blood sugar(148.3mg/dl)in cases of renal cell carcinoma along with diabetes mellitus was found in tumour size of 1-5 cm. CONCLUSION: Diabetes mellitus has independent prognostic significance in RCC in relation to tumour size and grade.

Association of type II diabetes mellitus with hepatocellular carcinoma occurrence--a case control study from Kathmandu Valley.

OBJECTIVE: To assess associations of Type II DM with hepatocellular carcinoma occurrence in Nepal. MATERIALS AND METHODS: This case control study was carried out using data retrieved from the register maintained in the Department of Biochemistry of Nepalese Army Institute of Health Sciences between 1st January, 2012, and 31st August, 2012. The variables collected were age, gender, HbA1c. All biochemical parameters were analyzed in the Central Laboratory of our hospital by standard validated methods. One way ANOVA was used to examine the statistical significant difference between groups with the LSD post-hoc test for comparison of means of case groups. Odds ratios (OR) were calculated using simple logistic-regression analysis. RESULTS: Etiological factors for HCC were HBV, HCV, alcohol and cryptogenic cirrhosis. The highest age group belonged to the etiological category of HCV with a mean of 71.9 ± 3.6 (CI 69.3, 74.5) years and the lowest age group to the etiological category of HBV with 61.7 ± 5.3(CI 57.9, 65.5) years. The main imperative basis of HCC in present study was HCV (39.5%) and second most significant cause of HCC was alcohol (26%). Glycated hemoglobin was found to be more in males with HCC (7.9%) as compared to females (7.3%). The percentage of Type II diabetes mellitus was greater in HCC patients when compared to controls. This difference was statistically significant with an odd ratio of 4.63 (p<0.001). CONCLUSION: Type II DM influences incidence, risk of recurrence, overall survival, and treatment-related complications in HCC patients.