A deep learning system for predicting time to progression of diabetic retinopathy.

Diabetic retinopathy (DR) is the leading cause of preventable blindness worldwide. The risk of DR progression is highly variable among different individuals, making it difficult to predict risk and personalize screening intervals. We developed and validated a deep learning system (DeepDR Plus) to predict time to DR progression within 5 years solely from fundus images. First, we used 717,308 fundus images from 179,327 participants with diabetes to pretrain the system. Subsequently, we trained and validated the system with a multiethnic dataset comprising 118,868 images from 29,868 participants with diabetes. For predicting time to DR progression, the system achieved concordance indexes of 0.754-0.846 and integrated Brier scores of 0.153-0.241 for all times up to 5 years. Furthermore, we validated the system in real-world cohorts of participants with diabetes. The integration with clinical workflow could potentially extend the mean screening interval from 12 months to 31.97 months, and the percentage of participants recommended to be screened at 1-5 years was 30.62%, 20.00%, 19.63%, 11.85% and 17.89%, respectively, while delayed detection of progression to vision-threatening DR was 0.18%. Altogether, the DeepDR Plus system could predict individualized risk and time to DR progression over 5 years, potentially allowing personalized screening intervals.

Mesenteric creeping fat index defined by CT enterography is associated with early postoperative recurrence in patients with Crohn's disease.

OBJECTIVES: To investigate the value of mesenteric creeping fat index (MCFI) defined by computed-tomography enterography (CTE) in patients with Crohn's Disease (CD) for predicting early postoperative recurrence. METHODS: A total of 110 patients with CD who underwent CTE and I-stage intestinal resection surgery from December 2013 to December 2018 were enrolled. Two radiologists independently assessed CTE parameters, including MCFI, with scores ranging from 1 to 8; bowel-wall thickening, with a scale of 1 to 3; mural hyperenhancement, mural stratification, submucosal fat deposition, mesenteric fibrofatty proliferation, mesenteric hypervascularity, mesenteric fat stranding, with a scale of 0 to 2; abscess/fistula, enlarged mesenteric lymph node, abdominal and pelvic effusion, with a scale of 0 to 1. Imaging findings associated with early recurrence were assessed using logistic regression analysis. RESULTS: Within one year follow-up, early postoperative recurrence occurred in 56.4 % (62/110) patients with CD. In univariate analysis, MCFI, bowel-wall thickening, mesenteric hypervascularity, mesenteric fat stranding, abscess/fistula and mesenteric lymphadenopathy were associated with early postoperative recurrence. Among all variables, MCFI (score ≥ 4) contributes the optimal AUC (0.838 [0.758-0.919]), specificity (89.6 %), positive predictive value (90.7 %), accuracy (83.6 %), and risk ratio (OR = 32.42 [10.69-98.33], p < 0.001). In multivariate analysis, only MCFI was an independent predictor of early postoperative recurrence (OR = 25.71 [7.65-86.35], p < 0.001). CONCLUSION: CTE features are useful in predicting early postoperative recurrence in patients with CD, MCFI may be a valuable tool for clinical monitoring and follow-up.

eXnet: An Efficient Approach for EmotionRecognition in the Wild.

Facial expression recognition has been well studied for its great importance in the areasof human-computer interaction and social sciences. With the evolution of deep learning, therehave been significant advances in this area that also surpass human-level accuracy. Althoughthese methods have achieved good accuracy, they are still suffering from two constraints (high computational power and memory), which are incredibly critical for small hardware-constrained devices. To alleviate this issue, we propose a new Convolutional Neural Network (CNN) architecture eXnet (Expression Net) based on parallel feature extraction which surpasses current methodsin accuracy and contains a much smaller number of parameters (eXnet: 4.57 million, VGG19:14.72 million), making it more efficient and lightweight for real-time systems. Several moderndata augmentation techniques are applied for generalization of eXnet; these techniques improvethe accuracy of the network by overcoming the problem of overfitting while containing the same size. We provide an extensive evaluation of our network against key methods on Facial ExpressionRecognition 2013 (FER-2013), Extended Cohn-Kanade Dataset (CK+), and Real-world Affective Faces Database (RAF-DB) benchmark datasets. We also perform ablation evaluation to show the importance of different components of our architecture. To evaluate the efficiency of eXnet on embedded systems,we  deploy it on Raspberry Pi 4B. All these evaluations show the superiority of eXnet for emotionrecognition in the wild in terms of accuracy, the number of parameters, and size on disk.

Personalized Exposure Control Using Adaptive Metering and Reinforcement Learning.

We propose a reinforcement learning approach for real-time exposure control of a mobile camera that is personalizable. Our approach is based on Markov Decision Process (MDP). In the camera viewfinder or live preview mode, given the current frame, our system predicts the change in exposure so as to optimize the trade-off among image quality, fast convergence, and minimal temporal oscillation. We model the exposure prediction function as a fully convolutional neural network that can be trained through Gaussian policy gradient in an end-to-end fashion. As a result, our system can associate scene semantics with exposure values; it can also be extended to personalize the exposure adjustments for a user and device. We improve the learning performance by incorporating an adaptive metering module that links semantics with exposure. This adaptive metering module generalizes the conventional spot or matrix metering techniques. We validate our system using the MIT FiveK [1] and our own datasets captured using iPhone 7 and Google Pixel. Experimental results show that our system exhibits stable real-time behavior while improving visual quality compared to what is achieved through native camera control.

Cerebrospinal fluid protein dynamic driver network: At the crossroads of brain tumorigenesis.

To get a better understanding of the ongoing in situ environmental changes preceding the brain tumorigenesis, we assessed cerebrospinal fluid (CSF) proteome profile changes in a glioma rat model in which brain tumor invariably developed after a single in utero exposure to the neurocarcinogen ethylnitrosourea (ENU). Computationally, the CSF proteome profile dynamics during the tumorigenesis can be modeled as non-smooth or even abrupt state changes. Such brain tumor environment transition analysis, correlating the CSF composition changes with the development of early cellular hyperplasia, can reveal the pathogenesis process at network level during a time before the image detection of the tumors. In our controlled rat model study, matched ENU- and saline-exposed rats' CSF proteomics changes were quantified at approximately 30, 60, 90, 120, 150 days of age (P30, P60, P90, P120, P150). We applied our transition-based network entropy (TNE) method to compute the CSF proteome changes in the ENU rat model and test the hypothesis of the critical transition state prior to impending hyperplasia. Our analysis identified a dynamic driver network (DDN) of CSF proteins related with the emerging tumorigenesis progressing from the non-hyperplasia state. The DDN associated leading network CSF proteins can allow the early detection of such dynamics before the catastrophic shift to the clear clinical landmarks in gliomas. Future characterization of the critical transition state (P60) during the brain tumor progression may reveal the underlying pathophysiology to device novel therapeutics preventing tumor formation. More detailed method and information are accessible through our website at http://translationalmedicine.stanford.edu.

Quantum algorithms for biomolecular solutions of the satisfiability problem on a quantum machine.

In this paper, we demonstrate that the logic computation performed by the DNA-based algorithm for solving general cases of the satisfiability problem can be implemented more efficiently by our proposed quantum algorithm on the quantum machine proposed by Deutsch. To test our theory, we carry out a three-quantum bit nuclear magnetic resonance experiment for solving the simplest satisfiability problem.

Fast parallel molecular algorithms for DNA-based computation: factoring integers.

The RSA public-key cryptosystem is an algorithm that converts input data to an unrecognizable encryption and converts the unrecognizable data back into its original decryption form. The security of the RSA public-key cryptosystem is based on the difficulty of factoring the product of two large prime numbers. This paper demonstrates to factor the product of two large prime numbers, and is a breakthrough in basic biological operations using a molecular computer. In order to achieve this, we propose three DNA-based algorithms for parallel subtractor, parallel comparator, and parallel modular arithmetic that formally verify our designed molecular solutions for factoring the product of two large prime numbers. Furthermore, this work indicates that the cryptosystems using public-key are perhaps insecure and also presents clear evidence of the ability of molecular computing to perform complicated mathematical operations.

Is optimal solution of every NP-complete or NP-hard problem determined from its characteristic for DNA-based computing.

Cook's Theorem [Cormen, T.H., Leiserson, C.E., Rivest, R.L., 2001. Introduction to Algorithms, second ed., The MIT Press; Garey, M.R., Johnson, D.S., 1979. Computer and Intractability, Freeman, San Fransico, CA] is that if one algorithm for an NP-complete or an NP-hard problem will be developed, then other problems will be solved by means of reduction to that problem. Cook's Theorem has been demonstrated to be correct in a general digital electronic computer. In this paper, we first propose a DNA algorithm for solving the vertex-cover problem. Then, we demonstrate that if the size of a reduced NP-complete or NP-hard problem is equal to or less than that of the vertex-cover problem, then the proposed algorithm can be directly used for solving the reduced NP-complete or NP-hard problem and Cook's Theorem is correct on DNA-based computing. Otherwise, a new DNA algorithm for optimal solution of a reduced NP-complete problem or a reduced NP-hard problem should be developed from the characteristic of NP-complete problems or NP-hard problems.

Molecular solutions for the subset-sum problem on DNA-based supercomputing.

In this paper our main purpose is to give molecular solutions for the subset-sum problem. In order to achieve this, we propose a DNA-based algorithm of an n-bit parallel adder and a DNA-based algorithm of an n-bit parallel comparator to formally verify our designed molecular solutions for the subset-sum problem.

Solving the set cover problem and the problem of exact cover by 3-sets in the Adleman-Lipton model.

Adleman wrote the first paper in which it is shown that deoxyribonucleic acid (DNA) strands could be employed towards calculating solutions to an instance of the NP-complete Hamiltonian path problem (HPP). Lipton also demonstrated that Adleman's techniques could be used to solve the NP-complete satisfiability (SAT) problem (the first NP-complete problem). In this paper, it is proved how the DNA operations presented by Adleman and Lipton can be used for developing DNA algorithms to resolving the set cover problem and the problem of exact cover by 3-sets.