Searcharxiv⌕ Search

arXiv subjects

Marleen Balvert

Publications and source records attributed to Marleen Balvert.

7 recordsLinked to original sources
arXiv2021

Iterative Rule Extension for Logic Analysis of Data: an MILP-based heuristic to derive interpretable binary classification from large datasets

Data-driven decision making is rapidly gaining popularity, fueled by the ever-increasing amounts of available data and encouraged by the development of models that can identify beyond linear input-output relationships. Simultaneously the need for interpretable prediction- and classification methods is increasing, as this improves both our trust in these models and the amount of information we can abstract from data. An important aspect of this interpretability is to obtain insight in the sensitivity-specificity trade-off constituted by multiple plausible input-output relationships. These are often shown in a receiver operating characteristic (ROC) curve. These developments combined lead to the need for a method that can abstract complex yet interpretable input-output relationships from large data, i.e. data containing large numbers of samples and sample features. Boolean phrases in disjunctive normal form (DNF) are highly suitable for explaining non-linear input-output relationships in a comprehensible way. Mixed integer linear programming (MILP) can be used to abstract these Boolean phrases from binary data, though its computational complexity prohibits the analysis of large datasets. This work presents IRELAND, an algorithm that allows for abstracting Boolean phrases in DNF from data with up to 10,000 samples and sample characteristics. The results show that for large datasets IRELAND outperforms the current state-of-the-art and can find solutions for datasets where current models run out of memory or need excessive runtimes. Additionally, by construction IRELAND allows for an efficient computation of the sensitivity-specificity trade-off curve, allowing for further understanding of the underlying input-output relationship.

cs.LG↗
arXiv2019

A Drug Recommendation System (Dr.S) for cancer cell lines

Personalizing drug prescriptions in cancer care based on genomic information requires associating genomic markers with treatment effects. This is an unsolved challenge requiring genomic patient data in yet unavailable volumes as well as appropriate quantitative methods. We attempt to solve this challenge for an experimental proxy for which sufficient data is available: 42 drugs tested on 1018 cancer cell lines. Our goal is to develop a method to identify the drug that is most promising based on a cell line's genomic information. For this, we need to identify for each drug the machine learning method, choice of hyperparameters and genomic features for optimal predictive performance. We extensively compare combinations of gene sets (both curated and random), genetic features, and machine learning algorithms for all 42 drugs. For each drug, the best performing combination (considering only the curated gene sets) is selected. We use these top model parameters for each drug to build and demonstrate a Drug Recommendation System (Dr.S). Insights resulting from this analysis are formulated as best practices for developing drug recommendation systems. The complete software system, called the Cell Line Analyzer, is written in Python and available on github.

stat.ML↗
arXiv2019

Dynamic fluence map sequencing using piecewise linear leaf position functions

Within the setting of intensity modulated radiation therapy (IMRT) and the fully continuous version of IMRT called volumetric modulated radiation therapy (VMAT), we consider the problem of matching a given fluence map as well as possible in limited time by the use of a linear accelerator (linac) with a multi-leaf collimator (MLC). We introduce two modeling strategies to manage the nonconvexity and the associated local minima of this problem. The first is the use of linear splines to model the MLC leaf positions as functions of time. The second is a progressively controllable smooth model (instead of a step function) of how the leaves block the fluence radiation. We propose a two part solution: an outer loop that optimizes the dose rate pattern over time, and an inner loop that given a dose rate pattern optimizes the leaf trajectories.

physics.med-ph↗
arXiv2018

An image representation based convolutional network for DNA classification

The folding structure of the DNA molecule combined with helper molecules, also referred to as the chromatin, is highly relevant for the functional properties of DNA. The chromatin structure is largely determined by the underlying primary DNA sequence, though the interaction is not yet fully understood. In this paper we develop a convolutional neural network that takes an image-representation of primary DNA sequence as its input, and predicts key determinants of chromatin structure. The method is developed such that it is capable of detecting interactions between distal elements in the DNA sequence, which are known to be highly relevant. Our experiments show that the method outperforms several existing methods both in terms of prediction accuracy and training time.

cs.LG↗
arXiv2016

Fast approximate delivery of fluence maps: the VMAT case

In this article we provide a method to generate the trade-off between delivery time and fluence map matching quality for volumetric modulated arc therapy (VMAT). At the heart of our method lies a mathematical programming model that, for a given duration of delivery, optimizes leaf trajectories and dose rates such that the desired fluence map is reproduced as well as possible. This model was presented for the single map case in a companion paper (Fast approximate delivery of fluence maps: the single map case). The resulting large-scale, non-convex optimization problem was solved using a heuristic approach. The single-map approach cannot directly be applied to the full arc case due to the large increase in model size, the issue of allocating delivery times to each of the arc segments, and the fact that the ending leaf positions for one map will be the starting leaf positions for the next map. In this article the method proposed in \cite{dm1} is extended to solve the full map treatment planning problem. We test our method using a prostate case and a head and neck case, and present the resulting trade-off curves. Analysis of the leaf trajectories reveal that short time plans have larger leaf openings in general than longer delivery time plans. Our method allows one to explore the continuum of possibilities between coarse, large segment plans characteristic of direct aperture approaches and narrow field plans produced by sliding window approaches. Exposing this trade off will allow for an informed choice between plan quality and solution time. Further research is required to speed up the optimization process to make this method clinically implementable.

physics.med-ph↗
arXiv2016

Fast approximate delivery of fluence maps: the single map case

In this first paper of a two-paper series, we present a method for optimizing the dynamic delivery of fluence maps in radiation therapy. For a given fluence map and a given delivery time, the optimization of the leaf trajectories of a multi-leaf collimator to approximately form the given fluence map is a non-convex optimization problem. Its general solution has not been addressed in the literature, despite the fact that dynamic delivery of fluence maps has long been a common approach to intensity modulated radiation therapy. We model the leaf trajectory and dose rate optimization as a non-convex continuous optimization problem and solve it by an interior point method from randomly initialized feasible starting solutions. We demonstrate the method on a fluence map from a prostate case and a larger fluence map from a head-and-neck case. While useful for static beam IMRT delivery, our main motivation for this work is the extension to the case of sequential fluence map delivery, i.e. the case of VMAT, which is the topic of the second paper.

physics.med-ph↗