arXiv subjects

Anirban Mukhopadhyay

Publications and source records attributed to Anirban Mukhopadhyay.

At least 91 records · Page 5Linked to original sources

arXiv2020

Predicting potential drug targets and repurposable drugs for COVID-19 via a deep generative model for graphs

Coronavirus Disease 2019 (COVID-19) has been creating a worldwide pandemic situation. Repurposing drugs, already shown to be free of harmful side effects, for the treatment of COVID-19 patients is an important option in launching novel therapeutic strategies. Therefore, reliable molecule interaction data are a crucial basis, where drug-/protein-protein interaction networks establish invaluable, year-long carefully curated data resources. However, these resources have not yet been systematically exploited using high-performance artificial intelligence approaches. Here, we combine three networks, two of which are year-long curated, and one of which, on SARS-CoV-2-human host-virus protein interactions, was published only most recently (30th of April 2020), raising a novel network that puts drugs, human and virus proteins into mutual context. We apply Variational Graph AutoEncoders (VGAEs), representing most advanced deep learning based methodology for the analysis of data that are subject to network constraints. Reliable simulations confirm that we operate at utmost accuracy in terms of predicting missing links. We then predict hitherto unknown links between drugs and human proteins against which virus proteins preferably bind. The corresponding therapeutic agents present splendid starting points for exploring novel host-directed therapy (HDT) options.

q-bio.MN↗

arXiv2020

M3d-CAM: A PyTorch library to generate 3D data attention maps for medical deep learning

M3d-CAM is an easy to use library for generating attention maps of CNN-based PyTorch models improving the interpretability of model predictions for humans. The attention maps can be generated with multiple methods like Guided Backpropagation, Grad-CAM, Guided Grad-CAM and Grad-CAM++. These attention maps visualize the regions in the input data that influenced the model prediction the most at a certain layer. Furthermore, M3d-CAM supports 2D and 3D data for the task of classification as well as for segmentation. A key feature is also that in most cases only a single line of code is required for generating attention maps for a model making M3d-CAM basically plug and play.

cs.CV↗

arXiv2020

AutoSNAP: Automatically Learning Neural Architectures for Instrument Pose Estimation

Despite recent successes, the advances in Deep Learning have not yet been fully translated to Computer Assisted Intervention (CAI) problems such as pose estimation of surgical instruments. Currently, neural architectures for classification and segmentation tasks are adopted ignoring significant discrepancies between CAI and these tasks. We propose an automatic framework (AutoSNAP) for instrument pose estimation problems, which discovers and learns the architectures for neural networks. We introduce 1)~an efficient testing environment for pose estimation, 2)~a powerful architecture representation based on novel Symbolic Neural Architecture Patterns (SNAPs), and 3)~an optimization of the architecture using an efficient search scheme. Using AutoSNAP, we discover an improved architecture (SNAPNet) which outperforms both the hand-engineered i3PosNet and the state-of-the-art architecture search method DARTS.

cs.CV↗

arXiv2020

Leveraging Spatial Uncertainty for Online Error Compensation in EMT

Purpose: Electromagnetic Tracking (EMT) can potentially complement fluoroscopic navigation, reducing radiation exposure in a hybrid setting. Due to the susceptibility to external distortions, systematic error in EMT needs to be compensated algorithmically. Compensation algorithms for EMT in guidewire procedures are only practical in an online setting. Methods: We collect positional data and train a symmetric Artificial Neural Network (ANN) architecture for compensating navigation error. The results are evaluated in both online and offline scenarios and are compared to polynomial fits. We assess spatial uncertainty of the compensation proposed by the ANN. Simulations based on real data show how this uncertainty measure can be utilized to improve accuracy and limit radiation exposure in hybrid navigation. Results: ANNs compensate unseen distortions by more than 70%, outperforming polynomial regression. Working on known distortions, ANNs outperform polynomials as well. We empirically demonstrate a linear relationship between tracking accuracy and model uncertainty. The effectiveness of hybrid tracking is shown in a simulation experiment. Conclusion: ANNs are suitable for EMT error compensation and can generalize across unseen distortions. Model uncertainty needs to be assessed when spatial error compensation algorithms are developed, so that training data collection can be optimized. Finally, we find that error compensation in EMT reduces the need for x-ray images in hybrid navigation.

eess.IV↗

arXiv2020

i3PosNet: Instrument Pose Estimation from X-Ray in temporal bone surgery

Purpose: Accurate estimation of the position and orientation (pose) of surgical instruments is crucial for delicate minimally invasive temporal bone surgery. Current techniques lack in accuracy and/or line-of-sight constraints (conventional tracking systems) or expose the patient to prohibitive ionizing radiation (intra-operative CT). A possible solution is to capture the instrument with a c-arm at irregular intervals and recover the pose from the image. Methods: i3PosNet infers the position and orientation of instruments from images using a pose estimation network. Said framework considers localized patches and outputs pseudo-landmarks. The pose is reconstructed from pseudo-landmarks by geometric considerations. Results: We show i3PosNet reaches errors less than 0.05mm. It outperforms conventional image registration-based approaches reducing average and maximum errors by at least two thirds. i3PosNet trained on synthetic images generalizes to real x-rays without any further adaptation. Conclusion: The translation of Deep Learning based methods to surgical applications is difficult, because large representative datasets for training and testing are not available. This work empirically shows sub-millimeter pose estimation trained solely based on synthetic training data.

cs.CV↗

arXiv2019

GANs for Medical Image Analysis

Generative Adversarial Networks (GANs) and their extensions have carved open many exciting ways to tackle well known and challenging medical image analysis problems such as medical image de-noising, reconstruction, segmentation, data simulation, detection or classification. Furthermore, their ability to synthesize images at unprecedented levels of realism also gives hope that the chronic scarcity of labeled data in the medical field can be resolved with the help of these generative models. In this review paper, a broad overview of recent literature on GANs for medical applications is given, the shortcomings and opportunities of the proposed methods are thoroughly discussed and potential future work is elaborated. We review the most relevant papers published until the submission date. For quick access, important details such as the underlying method, datasets and performance are tabulated. An interactive visualization which categorizes all papers to keep the review alive, is available at http://livingreview.in.tum.de/GANs_for_Medical_Applications.

cs.CV↗

arXiv2018

Exploring Adversarial Examples: Patterns of One-Pixel Attacks

Failure cases of black-box deep learning, e.g. adversarial examples, might have severe consequences in healthcare. Yet such failures are mostly studied in the context of real-world images with calibrated attacks. To demystify the adversarial examples, rigorous studies need to be designed. Unfortunately, complexity of the medical images hinders such study design directly from the medical images. We hypothesize that adversarial examples might result from the incorrect mapping of image space to the low dimensional generation manifold by deep networks. To test the hypothesis, we simplify a complex medical problem namely pose estimation of surgical tools into its barest form. An analytical decision boundary and exhaustive search of the one-pixel attack across multiple image dimensions let us localize the regions of frequent successful one-pixel attacks at the image space.

cs.CV↗

arXiv2018

Deep Spectral Correspondence for Matching Disparate Image Pairs

A novel, non-learning-based, saliency-aware, shape-cognizant correspondence determination technique is proposed for matching image pairs that are significantly disparate in nature. Images in the real world often exhibit high degrees of variation in scale, orientation, viewpoint, illumination and affine projection parameters, and are often accompanied by the presence of textureless regions and complete or partial occlusion of scene objects. The above conditions confound most correspondence determination techniques by rendering impractical the use of global contour-based descriptors or local pixel-level features for establishing correspondence. The proposed deep spectral correspondence (DSC) determination scheme harnesses the representational power of local feature descriptors to derive a complex high-level global shape representation for matching disparate images. The proposed scheme reasons about correspondence between disparate images using high-level global shape cues derived from low-level local feature descriptors. Consequently, the proposed scheme enjoys the best of both worlds, i.e., a high degree of invariance to affine parameters such as scale, orientation, viewpoint, illumination afforded by the global shape cues and robustness to occlusion provided by the low-level feature descriptors. While the shape-based component within the proposed scheme infers what to look for, an additional saliency-based component dictates where to look at thereby tackling the noisy correspondences arising from the presence of textureless regions and complex backgrounds. In the proposed scheme, a joint image graph is constructed using distances computed between interest points in the appearance (i.e., image) space. Eigenspectral decomposition of the joint image graph allows for reasoning about shape similarity to be performed jointly, in the appearance space and eigenspace.

cs.CV↗

arXiv2018

Measure Theoretic Aspects of Oscillations of Error Terms

We consider fluctuations of error terms $Δ(x)$ appearing in the asymptotic formula for a summatory function of coefficients of the Dirichlet series. These are quantified via $Ω$ and $Ω_{\pm}$ estimates. We obtain $Ω$ bounds for Lebesgue measure of the sets $$\{T\leq x \leq 2T: Δ(x)>λx^α\} \text{ and } \{T\leq x \leq 2T: Δ(x)< -λx^α\}$$ for some $α, λ>0$. Primary aim of this article is to develop a general framework to approach these problems. We rediscover several classical results in general setting with weak assumptions. Moreover, several applications of these methods have been discussed and new results have been obtained for some Dirichlet series.

math.NT↗

arXiv2018

Omega Theorems for The Twisted Divisor Function

For a fixed $θ\neq 0$, we define the twisted divisor function $$ τ(n, θ):=\sum_{d\mid n}d^{iθ}\ .$$ In this article we consider the error term $Δ(x)$ in the following asymptotic formula $$ \sum_{n\leq x}^*|τ(n, θ)|^2=ω_1(θ)x\log x + ω_2(θ)x\cos(θ\log x) +ω_3(θ)x + Δ(x),$$ where $ω_i(θ)$ for $i=1, 2, 3$ are constants depending only on $θ$. We obtain $$Δ(T)=Ω\left(T^{α(T)}\right) \text{ where } α(T) =\frac{3}{8}-\frac{c}{(\log T)^{1/8}} \text{ and } c>0,$$ along with an $Ω$-bound for the Lebesgue measure of the set of points where the above estimate holds.

math.NT↗

arXiv2018

A Theory of Diagnostic Interpretation in Supervised Classification

Interpretable deep learning is a fundamental building block towards safer AI, especially when the deployment possibilities of deep learning-based computer-aided medical diagnostic systems are so eminent. However, without a computational formulation of black-box interpretation, general interpretability research rely heavily on subjective bias. Clear decision structure of the medical diagnostics lets us approximate the decision process of a radiologist as a model - removed from subjective bias. We define the process of interpretation as a finite communication between a known model and a black-box model to optimally map the black box's decision process in the known model. Consequently, we define interpretability as maximal information gain over the initial uncertainty about the black-box's decision within finite communication. We relax this definition based on the observation that diagnostic interpretation is typically achieved by a process of minimal querying. We derive an algorithm to calculate diagnostic interpretability. The usual question of accuracy-interpretability tradeoff, i.e. whether a black-box model's prediction accuracy is dependent on its ability to be interpreted by a known source model, does not arise in this theory. With multiple example simulation experiments of various complexity levels, we demonstrate the working of such a theoretical model in synthetic supervised classification scenarios.

cs.LG↗

arXiv2018

How Bad is Good enough: Noisy annotations for instrument pose estimation

Though analysis of Medical Images by Deep Learning achieves unprecedented results across various applications, the effect of \emph{noisy training annotations} is rarely studied in a systematic manner. In Medical Image Analysis, most reports addressing this question concentrate on studying segmentation performance of deep learning classifiers. The absence of continuous ground truth annotations in these studies limits the value of conclusions for applications, where regression is the primary method of choice. In the application of surgical instrument pose estimation, where precision has a direct clinical impact on patient outcome, studying the effect of \emph{noisy annotations} on deep learning pose estimation techniques is of supreme importance. Real x-ray images are inadequate for this evaluation due to the unavailability of ground truth annotations. We circumvent this problem by generating synthetic radiographs, where the ground truth pose is known and therefore the pose estimation error made by the medical-expert can be estimated from experiments. Furthermore, this study shows the property of deep neural networks to learn dominant signals from noisy annotations with sufficient data in a regression setting.

cs.CV↗

arXiv2018

Bombieri-type theorem for convolution of arithmetic functions on Number field

Let $K$ be an imaginary quadratic number field of class number one and $\mathcal{O}_K$ be its ring of integers. We show that, if the arithmetic functions $f, g:\mathcal{O}_K\rightarrow \mathbb{C}$ both have level of distribution $\vartheta$ for some $0<\vartheta\leq 1/2$ then the Dirichlet convolution $f*g$ also have level of distribution $\vartheta$.

math.NT↗

arXiv2018

Planning Nonlinear Access Paths for Temporal Bone Surgery

Purpose: Interventions at the otobasis operate in the narrow region of the temporal bone where several highly sensitive organs define obstacles with minimal clearance for surgical instruments. Nonlinear trajectories for potential minimally-invasive interventions can provide larger distances to risk structures and optimized orientations of surgical instruments, thus improving clinical outcomes when compared to existing linear approaches. In this paper, we present fast and accurate planning methods for such nonlinear access paths. Methods: We define a specific motion planning problem in SE(3) = R3 x SO(3) with notable constraints in computation time and goal pose that reflect the requirements of temporal bone surgery.We then present k-RRT-Connect: two suitable motion planners based on bidirectional Rapidly-exploring Random Trees (RRT) to solve this problem efficiently. Results: The benefits of k-RRT-Connect are demonstrated on real CT data of patients. Their general performance is shown on a large set of realistic synthetic anatomies. We also show that these new algorithms outperform state of the art methods based on circular arcs or Bezier-Splines when applied to this specific problem. Conclusion: With this work we demonstrate that pre- and intra-operative planning of nonlinear access paths is possible for minimally-invasive surgeries at the otobasis.

cs.RO↗

arXiv2017

Quality Enhancement by Weighted Rank Aggregation of Crowd Opinion

Expertise of annotators has a major role in crowdsourcing based opinion aggregation models. In such frameworks, accuracy and biasness of annotators are occasionally taken as important features and based on them priority of the annotators are assigned. But instead of relying on a single feature, multiple features can be considered and separate rankings can be produced to judge the annotators properly. Finally, the aggregation of those rankings with perfect weightage can be done with an aim to produce better ground truth prediction. Here, we propose a novel weighted rank aggregation method and its efficacy with respect to other existing approaches is shown on artificial dataset. The effectiveness of weighted rank aggregation to enhance quality prediction is also shown by applying it on an Amazon Mechanical Turk (AMT) dataset.

cs.HC↗

arXiv2016

Tool and Phase recognition using contextual CNN features

A transfer learning method for generating features suitable for surgical tools and phase recognition from the ImageNet classification features [1] is proposed here. In addition, methods are developed for generating contextual features and combining them with time series analysis for final classification using multi-class random forest. The proposed pipeline is tested over the training and testing datasets of M2CAI16 challenges: tool and phase detection. Encouraging results are obtained by leave-one-out cross validation evaluation on the training dataset.

cs.CV↗

arXiv2016

Consensus of Dependent Opinions

Providing opinions through labeling of images, tweets, etc. have drawn immense interest in crowdsourcing markets. This invokes a major challenge of aggregating multiple opinions received from different crowd workers for deriving the final judgment. Generally, opinion aggregation models deal with independent opinions, which are given unanimously and are not visible to all. However, in many real-life cases, it is required to make the opinions public as soon as they are received. This makes the opinions dependent and might incorporate some bias. In this paper, we address a novel problem, hereafter denoted as dependent judgment analysis, and discuss the requirements for developing an appropriate model to deal with this problem. The challenge remains to be improving the consensus by revealing true opinions.

cs.HC↗

arXiv2016

A Markov Chain based Ensemble Method for Crowdsourced Clustering

In presence of multiple clustering solutions for the same dataset, a clustering ensemble approach aims to yield a single clustering of the dataset by achieving a consensus among the input clustering solutions. The goal of this consensus is to improve the quality of clustering. It has been seen that there are some image clustering tasks that cannot be easily solved by computer. But if these images can be outsourced to the general people (crowd workers) to group them based on some similar features, and opinions are collected from them, then this task can be managed in an efficient manner and time effective way. In this work, the power of crowd has been used to annotate the images so that multiple clustering solutions can be obtained from them and thereafter a Markov chain based ensemble method is introduced to make a consensus of multiple clustering solutions.

cs.HC↗