Receive a weekly summary and discussion of the top papers of the week by leading researchers in the field.

Radiology Radiology

Deciphering multiple sclerosis disability with deep learning attention maps on clinical MRI.

In NeuroImage. Clinical

The application of convolutional neural networks (CNNs) to MRI data has emerged as a promising approach to achieving unprecedented levels of accuracy when predicting the course of neurological conditions, including multiple sclerosis, by means of extracting image features not detectable through conventional methods. Additionally, the study of CNN-derived attention maps, which indicate the most relevant anatomical features for CNN-based decisions, has the potential to uncover key disease mechanisms leading to disability accumulation. From a cohort of patients prospectively followed up after a first demyelinating attack, we selected those with T1-weighted and T2-FLAIR brain MRI sequences available for image analysis and a clinical assessment performed within the following six months (N = 319). Patients were divided into two groups according to expanded disability status scale (EDSS) score: ≥3.0 and < 3.0. A 3D-CNN model predicted the class using whole-brain MRI scans as input. A comparison with a logistic regression (LR) model using volumetric measurements as explanatory variables and a validation of the CNN model on an independent dataset with similar characteristics (N = 440) were also performed. The layer-wise relevance propagation method was used to obtain individual attention maps. The CNN model achieved a mean accuracy of 79% and proved to be superior to the equivalent LR-model (77%). Additionally, the model was successfully validated in the independent external cohort without any re-training (accuracy = 71%). Attention-map analyses revealed the predominant role of frontotemporal cortex and cerebellum for CNN decisions, suggesting that the mechanisms leading to disability accrual exceed the mere presence of brain lesions or atrophy and probably involve how damage is distributed in the central nervous system.

Coll Llucia, Pareto Deborah, Carbonell-Mirabent Pere, Cobo-Calvo Álvaro, Arrambide Georgina, Vidal-Jordana Ángela, Comabella Manuel, Castilló Joaquín, Rodríguez-Acevedo Breogán, Zabalza Ana, Galán Ingrid, Midaglia Luciana, Nos Carlos, Salerno Annalaura, Auger Cristina, Alberich Manel, Río Jordi, Sastre-Garriga Jaume, Oliver Arnau, Montalban Xavier, Rovira Àlex, Tintoré Mar, Lladó Xavier, Tur Carmen

2023-Mar-15

Attention maps, Deep learning, Disability, Multiple sclerosis, Structural MRI

General General

Facial reconstruction using 3-D computerized method: A scoping review of Methods, current Status, and future developments.

In Legal medicine (Tokyo, Japan)

Facial reconstruction (otherwise known as facial approximation) is an alternative method that has been widely accepted in forensic anthropological and archaeological circumstances. This method is considered useful for creating the virtual face of a person based on skull remain. Three-dimensional (3-D) traditional facial reconstruction (known as sculpture or manual method) has been recognized for more than a century; however, it was declared to be subjective and required anthropological training. Until recently, with the progression of computational technologies, many studies attempted to develop a more appropriate method, so-called the 3-D computerized facial reconstruction. This method also relied on anatomical knowledge of the face-skull relationship, divided into semi- and automated based computational method. The 3-D computerized facial reconstruction makes it more rapid, more flexible, and more realistic to generate multiple representations of faces. Moreover, new tools and technology are continuously generating fascinating and sound research as well as encouraging multidisciplinary collaboration. This has led to a paradigm shift in the 3-D computerized facial reconstruction to a new finding and new technique based on artificial intelligence in academia. Based on the last 10-years scientific-published documents, this article aims to explain the overview of the 3-D computerized facial reconstruction and progression as well as an issue relating to future directions to encourage further improvement.

Navic Pagorn, Inthasan Chanatporn, Chaimongkhol Thawanthorn, Mahakkanukrauh Pasuk

2023-Mar-15

Computer-assisted, Facial approximation, Facial reconstruction, Forensic anthropology, Review, Three-dimensional

Internal Medicine Internal Medicine

A New Tool for Holistic Residency Application Review: Using Natural Language Processing of Applicant Experiences to Predict Interview Invitation.

In Academic medicine : journal of the Association of American Medical Colleges

PROBLEM : Reviewing residency application narrative components is time intensive and has contributed in part to nearly half of all applications not receiving holistic review. The authors developed a natural language processing (NLP) based tool to automate review of applicants' narrative experience entries and predict interview invitation.

APPROACH : Experience entries (n = 188,500) were extracted from 6,403 residency applications across 3 application cycles (2017-2019) at 1 internal medicine program, combined at the applicant level, and paired with the interview invitation decision (n = 1,224 invitations). NLP identified important words (or word pairs) with term frequency-inverse document frequency, which were used to predict interview invitation using logistic regression with L1 regularization. Terms remaining in the model were analyzed thematically. Logistic regression models were also built using structured application data and a combination of NLP and structured data. Model performance was evaluated on never-before-seen data using area under the receiver operating characteristic and precision-recall curves (AUROC, AUPRC).

OUTCOMES : The NLP model had an AUROC of 0.80 (vs. chance decision of 0.50) and AUPRC of 0.49 (vs. chance decision of 0.19), showing moderate predictive strength. Phrases indicating active leadership, research, or work in social justice and health disparities were associated with interview invitation. The model's detection of these key selection factors demonstrated face validity. Adding structured data to the model significantly improved prediction (AUROC 0.92, AUPRC 0.73), as expected given reliance on such metrics for interview invitation.

NEXT STEPS : This model represents a first step in using NLP-based artificial intelligence tools to promote holistic residency application review. The authors are assessing the practical utility of using this model to identify applicants screened out using traditional metrics. Generalizability must be determined through model retraining and evaluation at other programs. Work is ongoing to thwart model "gaming," improve prediction, and remove unwanted biases introduced during model training.

Mahtani Arun Umesh, Reinstein Ilan, Marin Marina, Burk-Rafel Jesse

2023-Mar-16

oncology Oncology

A pair of deep learning auto-contouring models for prostate cancer patients injected with a radio-transparent versus radiopaque hydrogel spacer.

In Medical physics ; h5-index 59.0

BACKGROUND : Absorbable hydrogel spacer injected between prostate and rectum is gaining popularity for rectal sparing. The spacer alters patient anatomy and thus requires new auto-contouring models.

PURPOSE : To report the development and comprehensive evaluation of two deep-learning models for patients injected with a radio-transparent (model I) versus radiopaque (model II) spacer.

METHODS AND MATERIALS : Model I was trained and cross-validated by 135 cases with transparent spacer and tested on 24 cases. Using refined training methods, model II was trained and cross-validated by the same dataset, but with the Hounsfield Unit distribution in the spacer overridden by that obtained from ten cases with opaque spacer. Model II was tested on 64 cases. The models auto-contour eight regions of interest (ROIs): spacer, prostate, proximal seminal vesicles (SVs), left and right femurs, bladder, rectum and penile bulb. Qualitatively, each auto contour (AC), as well as the composite set, was assessed against manual contour (MC), by a radiation oncologist using a 1 (accepted directly or after minor editing), 2 (accepted after moderate editing), 3 (accepted after major editing) and 4 (rejected) scoring scale. The efficiency gain was characterized by the mean score as nearly complete [1 to 1.75], substantial (1.75 to 2.5], meaningful (2.5 to 3.25] and no (3.25 to 4.00]. Quantitatively, the geometric similarity between AC and MC was evaluated by Dice similarity coefficient (DSC) and mean distance to agreement (MDA), using tolerance recommended by AAPM TG-132 Report. The results by the two models were compared to examine the outcome of the refined training methods. The large number of testing cases for model II allowed further investigation of inter-observer variability in clinical dataset. The correlation between score and DSC/MDA was studied on the ROIs with ten or more counts of each acceptable score (1, 2, 3).

RESULTS : For model I/model II: the mean score was 3.63/1.30 for transparent/opaque spacer, 2.71/2.16 for prostate, 3.25/2.44 for proximal SVs, 1.13/1.02 for both femurs, 2.25/1.25 for bladder, 3.00/2.06 for rectum, 3.38/2.42 for penile bulb and 2.79/2.20 for the composite set; the mean DSC was 0.52/0.84 for spacer, 0.84/0.85 for prostate, 0.60/0.62 for proximal SVs, 0.94/0.96 for left femur, 0.95/0.96 for right femur, 0.91/0.95 for bladder, 0.81/0.84 for rectum and 0.65/0.65 for penile bulb; and the mean MDA was 2.9/0.9 mm for spacer, 1.9/1.7 mm for prostate, 2.4/2.3 mm for proximal SVs, 0.8/0.5 mm for left femur, 0.7/0.5 mm for right femur, 1.5/0.9 mm for bladder, 2.3/1.9 mm for rectum and 2.2/2.2 mm for penile bulb. Model II showed significantly improved scores for all ROIs, and metrics for spacer, femurs, bladder and rectum. Significant inter-observer variability was only found for prostate. Highly linear correlation between the score and DSC was found for the two qualified ROIs (prostate and rectum).

CONCLUSIONS : The overall efficiency gain was meaningful for model I and substantial for model II. The ROIs meeting the clinical deployment criteria (mean score below 3.25, DSC above 0.8 and MDA below 2.5 mm) included prostate, both femurs, bladder and rectum for both models, and spacer for model II. This article is protected by copyright. All rights reserved.

Wang Yi, Boyd Graham, Zieminski Stephen, Kamran Sophia C, Zietman Anthony L, Miyamoto David T, Kirk Maxwell C, Efstathiou Jason A

2023-Mar-20

auto-contouring, deep learning, hydrogel rectal spacer, inter-observer variability, physician score, prostate radiotherapy, quantitative metrics

General General

Predicting Churn in Online Games by Quantifying Diversity of Engagement.

In Big data

Understanding engagement patterns of users in online platforms, may it be games, online social networks, or academic websites, is a widely studied topic with many real-world applications and economic consequences. A holy grail in this area of research is to develop an automatic prediction algorithm for when a user is going to leave the platform and devise proper intervention. In this work, we study online recreational games and propose to model the engagement patterns of players through an unsupervised learning framework. We think of engagement as a continuous temporal process, measured along specific axes derived from gaming users' data using principal component analysis. We track the overall trend of the projection of the data along the significant principal components. We find that the geometric variability of the trajectory is a good predictor of the users' engagement level. Users characterized by a time series with large variability are users with higher engagement; namely, they will continue playing the game for prolonged periods of time. We evaluated our methodology on two data sets of very different game types and compared the performance of our method with state-of-the-art black-box machine learning algorithms. Our results were fairly competitive with these methods, and we conclude that churn can be predicted using an explainable, intuitive, and white-box decision-rule algorithm.

Weiss Idan, Vilenchik Dan

2023-Mar-20

PCA, casual online gaming, churn prediction, disengagement, explainable AI, simple heuristics, unsupervised learning

General General

Metheor: Ultrafast DNA methylation heterogeneity calculation from bisulfite read alignments.

In PLoS computational biology

Phased DNA methylation states within bisulfite sequencing reads are valuable source of information that can be used to estimate epigenetic diversity across cells as well as epigenomic instability in individual cells. Various measures capturing the heterogeneity of DNA methylation states have been proposed for a decade. However, in routine analyses on DNA methylation, this heterogeneity is often ignored by computing average methylation levels at CpG sites, even though such information exists in bisulfite sequencing data in the form of phased methylation states, or methylation patterns. In this study, to facilitate the application of the DNA methylation heterogeneity measures in downstream epigenomic analyses, we present a Rust-based, extremely fast and lightweight bioinformatics toolkit called Metheor. As the analysis of DNA methylation heterogeneity requires the examination of pairs or groups of CpGs throughout the genome, existing softwares suffer from high computational burden, which almost make a large-scale DNA methylation heterogeneity studies intractable for researchers with limited resources. In this study, we benchmark the performance of Metheor against existing code implementations for DNA methylation heterogeneity measures in three different scenarios of simulated bisulfite sequencing datasets. Metheor was shown to dramatically reduce the execution time up to 300-fold and memory footprint up to 60-fold, while producing identical results with the original implementation, thereby facilitating a large-scale study of DNA methylation heterogeneity profiles. To demonstrate the utility of the low computational burden of Metheor, we show that the methylation heterogeneity profiles of 928 cancer cell lines can be computed with standard computing resources. With those profiles, we reveal the association between DNA methylation heterogeneity and various omics features. Source code for Metheor is at https://github.com/dohlee/metheor and is freely available under the GPL-3.0 license.

Lee Dohoon, Koo Bonil, Yang Jeewon, Kim Sun

2023-Mar-20