Bulletin of the American Physical Society
APS March Meeting 2022
Volume 67, Number 3
Monday–Friday, March 14–18, 2022; Chicago
Session Q29: Artificial Intelligence, Machine Learning, and Quantitative Biomarkers in Medicine and BiomedicineFocus Session Recordings Available
|
Hide Abstracts |
Chair: MIchael Boss, American College of Radiology Room: McCormick Place W-190B |
Wednesday, March 16, 2022 3:00PM - 3:36PM |
Q29.00001: The use of deep learning with neuroimaging for diagnosing and monitoring disease Invited Speaker: Alison Deatsch Despite clear clinical necessity, many neurodegenerative diseases lack comprehensive, generalizable tools for reliable diagnosis, progression monitoring, and personalized approaches to disease management. Quantitative medical imaging holds rich, patient-specific disease information which may address these clinical deficits, but it is rarely used to its full potential due to challenges with inherent uncertainties and time-consuming data curation. To address these challenges, the field is moving beyond qualitative, manual image interpretation towards a quantitative data science approach to patient management which considers both imaging and clinical data. Advanced computational techniques, particularly in deep learning, can detect hidden quantitative imaging features, efficiently maximize information output from neurological images, and synthesize this information with clinical data to detect, predict, and model disease. This appreciation for quantitative science is opening opportunities for medical physics to advance the understanding of neurodegenerative disease and enhance treatment decision-making in the clinic. |
Wednesday, March 16, 2022 3:36PM - 3:48PM |
Q29.00002: Uncertainty Estimation for Deep Learning-based image segmentation via Monte Carlo test-time dropout: Application on pectoral muscle segmentation from Full Field Digital Mammography images Zan Klanecek, Lesley Cockmartin, Kristijana Hertl, Daniel Huff, Katja Jarm, Mateja Krajc, Nicholas Marshall, Andrej Studen, Milos Vrhovec, Tobias Wagner, Yao Kuan Wang, Hilde Bosmans, Robert Jeraj Most deep-learning (DL) models used for image analysis lack interpretability and mechanisms to evaluate associated uncertainty. We addressed this challenge on a problem of pectoral muscle (PM) segmentation from Full Field Digital Mammography (FFDM) scans – an essential step for automated breast cancer risk prediction. |
Wednesday, March 16, 2022 3:48PM - 4:00PM |
Q29.00003: Analysis of skin hyperspectral images by machine learning methods Matija Milanic, Teo Manojlovic, Tadej Tomanic, Ivan Stajduhar Hyperspectral images (HSI) are optical images containing both spatial and spectral information about the imaged object. The current image analysis include different derivatives of the radiative transfer equation, which are either slow or inaccurate. A possible solution is implementation of machine learning methods (ML). |
Wednesday, March 16, 2022 4:00PM - 4:12PM |
Q29.00004: Improved Biexponential Decay Parameter Estimation from Input-Layer Regularization of a Neural Network Richard G Spencer, Jonathan L Palumbo, Michael Rozowski, Chuan Bi, Wojciech Czaja, Jay Bisen A decaying biexponential model is used to describe relaxation processes in biophysical systems as disparate as drug pharmacokinetics, fluorescence imaging, and magnetic resonance relaxometry (MRR). However, estimation of biexponential signal parameters is well-known to be an ill-posed problem, with distinct parameter sets producing nearly identical decay curves. Recently introduced Neural network (NN) approaches out-perform non-linear least squares (NLLS) analysis. Regularization is an important component of NN design to prevent overfitting; here, we introduce a new type of NN regularization based on Tikhonov regularization at the input layer. We achieve robust performance in signals with limited SNR by incorporating a larger region of parameter space into the training set than in the testing set. We find improvements in accuracy and precision of decay constant estimation on the order of 15% over conventional advanced NN analyses, and of roughly 40% over NLLS, highlighting the additional content provided by regularization. Input layer regularization is compatible with more conventional methods of NN regularization, such as epoch analysis and weight drop-out. In our main application of MRR, these results will improve macromolecular mapping in the brain and other tissues. |
Wednesday, March 16, 2022 4:12PM - 4:24PM |
Q29.00005: Breaking the CRLB Barrier: Decreasing Mean Squared Error in Parameter Estimation Through Introduction of Regularization Bias Richard G Spencer, Ryan Neff, Chuan Bi, Radu Balan, Zezheng Song The inverse problem of parameter estimation arises throughout physics; one important example is the decaying biexponential (BE) or similar models describing phenomena as disparate as fluorescence decay, supernova light curves, and magnetic resonance relaxometry (MRR). However, BE parameter estimation is highly ill-posed, with results unstable with respect to noise. Therefore, we sought to incorporate Tikhonov regularization into the estimation problem to decrease variance at the expense of introducing a smaller amount of bias in order to decrease mean square error (MSE). In the difficult problem of estimating the four parameters of the BE model, we find that for typical signal-to-noise and reasonable MRR transverse relaxation (T2) ranges, the optimal regularization parameter can decrease MSE by an order of magnitude or more as compared to non-regularized least squares. While the mathematically optimal cannot be determined for experimental data with unknown generative parameters, standard approaches to selection of , such as generalized cross-validation, reliably yield improvements in MSE on the order of 50%. We also find that the variance of the estimators is below the Cramer-Rao lower bound, applicable only to unbiased estimators. In our primary application of MRR, these results will improve macromolecular mapping in the brain and other tissues. |
Wednesday, March 16, 2022 4:24PM - 4:36PM |
Q29.00006: Radiomics based predication of radiation induced xerostomia for head and neck cancer patients Saad Bin Saeed Ahmed We proposed to develop a radiomics driven machine learning prediction model. The model is to predict radiation induced xerostomia in head and neck squamous cell carcinoma (HNSCC) patients receiving radiation therapy. The aim is to predict the xerostomia levels and help radiation oncologist to re-plan or revise the radiation treatment. We chose a dataset consist of 31 HNSCC patients who underwent highly conformal radiation therapy. Each patient had pre-, mid- and post-treatment scans with contoured planning structures. It also consists of clinical follow up measurements of patients. We divided the date-set into feature extraction, training, and validation categories. Radiomic features were extracted from parotids structure of each patient in all three CT images to quantify the changes occurred during different stages of radiation therapy. Six significant radiomic features i.e., NGLDM_Coarseness, GLRLM_LGRE, GLRLM_GLNU, NGLDM_Busyness, GLZLM_LGZE and GLRLM_SRE were found based on statistical correlation with clinical outcome. GLZLM_LGZE and GLRLM_SRE are found significant in both definitive and post-surgery cases. The selected radiomic features will be utilized during training phase of machine learning model as classifier. |
Wednesday, March 16, 2022 4:36PM - 4:48PM |
Q29.00007: On the use of Image Texture in Medical Data Science Mini Das, Diego Andrade Texture features provide statistical, spatial, and structural information of the pixel arrangement of a digital image. A considerable number of new cutting-edge advancements in the medical imaging field involve the usage or application of machine learning and AI techniques, and in many cases, it has been paired with image texture analysis to improve results. We will examine factors that effect the robustness of image texture feature estimations as well as specific features and estimations strategies that would aid segmentation and pattern analysis. Examples will be provided from both simulated and experimental data involving tomographic imaging methods such as partial angle breast tomosyntehsis, computed tomography (CT) and micro CT. |
Wednesday, March 16, 2022 4:48PM - 5:00PM |
Q29.00008: Disease related metabolic patterns as imaging biomarkers for multiple system atrophy and progressive supranuclear palsy Eva Rebec, Petra Tomse, Andrej Studen, Matej Perovnik, Urban Simoncic, Maja Trost Besides Parkinson’s disease, parkinsonisms comprises of atypical parkinsonisms - multiple system atrophy (MSA), progressive supranuclear palsy (PSP), and corticobasal degeneration. Distinguishing between them is difficult; therefore a biomarker is needed. |
Wednesday, March 16, 2022 5:00PM - 5:12PM |
Q29.00009: Effect of image resolution on the diagnostic performance of disease-related patterns derived from brain FDG-PET images Urban Simoncic, Tadej Tomanic, Eva Rebec, Maja Trost Dementia is often caused by Alzheimer’s disease (AD) but may have other causes. Finding the cause may be aided by characteristic metabolic brain patterns exracted from 18F-fluorodeoxyglucose positron emission tomography images. We investigated the effect of image resolution on the AD-related pattern’s (ADRP) diagnostic performance. |
Wednesday, March 16, 2022 5:12PM - 5:24PM |
Q29.00010: Improvement of a Radiomics Based Automated Breast Density Algorithm Evaluated on a Time Series of Mammograms Andrej Studen, Zan Klanecek, Mateja Krajc, Milos Vrhovec, Kristijana Hertl, Robert Jeraj, Katja Jarm, Luka Premoša, Jan Štefanič Radiographic breast density is an important independent breast cancer risk predictor. Current automated density estimation algorithms have a limited prediction power. Normally they analyze full field digital mammograms recorded at a single screening visit. We have developed innovative time-series image analytics and compared it to a single time-point prediction. We looked at 3200 cancer free participants of the Slovenian Breast Cancer Screening Programme DORA that had a mammogram that needed further screening assessment. Our prediction algorithm was optimized to the breast density scored by a radiologist as part of the assessment. We combined multiple predictions from the series using a soft voting classifier with specific weights optimized for accuracy. For the reference visit only, the Cohen kappa score was 0.63±0.01. Pairing them with earlier predictions, kappa grew to 0.67±0.01, and earlier predictions were favored with a weight of 0.55±0.05. With future predictions, no improvements were seen, and low weights were assigned. The accuracy of our density scoring algorithm with performance comparable to algorithms reported in literature, was improved slightly using a time-series of mammograms. The improvement relied on earlier, but not on mammograms recorded after the scoring visit. |
Wednesday, March 16, 2022 5:24PM - 5:36PM |
Q29.00011: Automatic Prescription Anomaly Detection Tool Assisting Peer Review Chart Rounds in Radiotherapy Qiongge Li, Jean Wright, Russell Hales, Todd McNutt, Ranh Voong Appropriate dosage of radiation is crucial in patient safety in radiotherapy. The current quality assurance heavily depends on a peer-review chart-round process, where the physicians reach a consensus on each patient’s dosage. However, such a process is manual and laborious. Physicians might not catch an error because of their limited time and energy. We designed a novel anomaly detection algorithm that utilized historical data from the past to predict anomalous cases. Such a tool can serve as an electronic peer who will assist the peer-review process providing extra safety to the patients. |
Wednesday, March 16, 2022 5:36PM - 5:48PM |
Q29.00012: Solution to a Mathematical Model of Cheyne-Stokes Breathing John B Delos Cheyne-Stokes Breathing (CSB) often occurs in patients with congestive heart failure and may be a prelude to death. A similar phenomenon is very common in healthy premature infants, but excessive CSB has been correlated with sudden infant death. It follows that mathematical theories of CSB are of great interest. Our goal is predictive monitoring of patients in intensive care: by applying mathematical-physiological theories to respiratory data, we hope to extract parameters (time-delays, feedback coefficients, loop gain, etc.), and then to seek patterns in the behavior of these parameters that occur prior to -- and are reasonably predictive of -- important clinical events. Here we develop and present solutions to a mathematical theory that may extract new information from respiratory signals. |
Wednesday, March 16, 2022 5:48PM - 6:00PM |
Q29.00013: Statistical parametric mapping on physiological maps of human hands with arthritic joints Luka Rogelj, Tadej Tomanič, Matija Milanic, Urban Simoncic Statistical parametric mapping (SPM) refers to the construction of spatially extended statistical processes to test the hypothesis of regional effects. It is commonly used to identify regionally specific effects in neuroimaging data to characterize brain activity and disease-related changes. |
Follow Us |
Engage
Become an APS Member |
My APS
Renew Membership |
Information for |
About APSThe American Physical Society (APS) is a non-profit membership organization working to advance the knowledge of physics. |
© 2025 American Physical Society
| All rights reserved | Terms of Use
| Contact Us
Headquarters
1 Physics Ellipse, College Park, MD 20740-3844
(301) 209-3200
Editorial Office
100 Motor Pkwy, Suite 110, Hauppauge, NY 11788
(631) 591-4000
Office of Public Affairs
529 14th St NW, Suite 1050, Washington, D.C. 20045-2001
(202) 662-8700