arXiv subjects

David Sterling

Publications and source records attributed to David Sterling.

4 recordsLinked to original sources

arXiv2025

Optimization models for needle placement in 3D-printed masks for high dose rate brachytherapy

High dose rate brachytherapy (HDR-BT) is an appealing treatment option for superficial cancers that permits the delivery of higher local doses than other radiation modalities without a significant increase in toxicity. In order for HDR-BT to be used in these situations, needles through which the radiation source is passed must be strategically placed in close proximity to the patient's body. Currently, this crucial step is performed manually by physicians or medical physicists. The use of 3D-printed masks customized for individual patients has been advocated as a way to more precisely and securely position these needles, with the potential of producing better and safer treatment plans. In this paper, we propose optimization approaches for positioning needles within 3D-printed masks for HDR-BT, focusing on skin cancers. We numerically show that the models we propose efficiently generate more homogeneous plans than those derived manually and provide an alternative to manual placement that can save planning time and enhance plan quality.

physics.med-ph↗

arXiv2020

Scalable Label Propagation for Multi-relational Learning on the Tensor Product of Graphs

Multi-relational learning on knowledge graphs infers high-order relations among the entities across the graphs. This learning task can be solved by label propagation on the tensor product of the knowledge graphs to learn the high-order relations as a tensor. In this paper, we generalize a widely used label propagation model to the normalized tensor product graph, and propose an optimization formulation and a scalable Low-rank Tensor-based Label Propagation algorithm (LowrankTLP) to infer multi-relations for two learning tasks, hyperlink prediction and multiple graph alignment. The optimization formulation minimizes the upper bound of the noisy tensor estimation error for multiple graph alignment, by learning with a subset of the eigen-pairs in the spectrum of the normalized tensor product graph. We also provide a data-dependent transductive Rademacher bound for binary hyperlink prediction. We accelerate LowrankTLP with parallel tensor computation which enables label propagation on a tensor product of 100 graphs each of size 1000 in less than half hour in the simulation. LowrankTLP was also applied to predicting the author-paper-venue hyperlinks in publication records, alignment of segmented regions across up to 26 CT-scan images and alignment of protein-protein interaction networks across multiple species. The experiments demonstrate that LowrankTLP indeed well approximates the original label propagation with better scalability and accuracy.

cs.LG↗

arXiv2016

Rising temperatures increased recruitment of brown tiger prawn (Penaeus esculentus) in Moreton Bay (Australia)

Abiotic factors are fundamental drivers of the dynamics of wild marine fish populations. Identifying and quantifying their influence on species targeted by the fishing industry is difficult and very important for managing fisheries in a changing climate. Using multiple regression, we investigated the influence of both temperature and rainfall on the variability of recruitment of a tropical species, the brown tiger prawn (Penaeus esculentus), in Moreton Bay which is located near the southern limit of its distribution on the east coast of Australia. A step-wise selection between environmental variables identified that variations in recruitment from 1990 to 2014 were best explained by a combination of temperature and spawning stock biomass. Temperature explains 35% of recruitment variability and spawning stock biomass 33%. This analysis suggests that increasing temperatures have increased recruitment of brown tiger prawn in Moreton Bay.

q-bio.PE↗

arXiv2015

A maximum likelihood estimate of natural mortality for brown tiger prawn (Penaeus esculentus) in Moreton Bay (Australia)

The delay difference model was implemented to fit 21 years of brown tiger prawn (Penaeus esculentus) catch in Moreton Bay by maximum likelihood to assess the status of this stock. Monte Carlo simulations testing of the stock assessment software coded in C++ showed that the model could estimate simultaneously natural mortality in addition to catchability, recruitment and initial biomasses. Applied to logbooks data collected from 1990 to 2010, this implementation of the delay difference provided for the first time an estimate of natural mortality for brown tiger prawn in Moreton Bay, equal to $0.031 \pm 0.002$ week$^{-1}$. This estimate is approximately 30\% lower than the value of natural mortality (0.045 week$^{-1}$) used in previous stock assessments of this species.

q-bio.QM↗