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

General General

3D deep learning model for the pretreatment evaluation of treatment response in locally advanced TESCC: A prospective study.

In International journal of radiation oncology, biology, physics

PURPOSE : To develop and validate a pretreatment CT-based deep-learning (DL) model for predicting the treatment response to concurrent chemoradiotherapy (CCRT) among patients with locally advanced thoracic esophageal squamous cell carcinoma (TESCC).

MATERIALS AND METHODS : We conducted a prospective, multicenter study on the therapeutic efficacy of CCRT among TESCC patients across 9 hospitals in China (ChiCTR**********). A total of 306 patients with locally advanced TESCC diagnosed by histopathology from August 2015 to May 2020 were included in this study. A 3D deep-learning radiomics model (3D-DLRM) was developed and validated based on pretreatment CT images to predict the response to CCRT. Furthermore, the prediction performance of the newly developed 3D-DLRM was analyzed according to 3 categories: radiotherapy plan, radiation field, and prescription dose used.

RESULTS : The 3D-DLRM achieved good prediction performance, with areas under the receiver operating characteristic curve (AUCs) of 0.897 (95% CI 0.840-0.959) for the training cohort and 0.833 (95% CI 0.654-1.000) for the validation cohort. Specifically, the 3D-DLRM accurately predicted patients who would not respond to CCRT, with a positive predictive value (PPV) of 100% for the validation cohort. Moreover, the 3D-DLRM performed well in all 3 categories, each with AUCs > 0.8 and PPVs of approximately 100%.

CONCLUSION : The proposed pretreatment CT-based 3D-DLRM provides a potential tool for predicting the response to CCRT among patients with locally advanced TESCC. With the help of precise pretreatment prediction, we may guide the individualized treatment of patients and improve survival.

Li Xiaoqin, Gao Han, Zhu Jian, Huang Yong, Zhu Yongbei, Huang Wei, Li Zhenjiang, Sun Kai, Liu Zhenyu, Tian Jie, Li Baosheng


General General

Roots are key to increasing the mean residence time of organic carbon entering temperate agricultural soils.

In Global change biology

The ratio of soil organic carbon stock (SOC) to annual carbon input gives an estimate of the mean residence time of organic carbon that enters the soil (MRTOC ). It indicates how efficiently biomass can be transformed into SOC, which is of particular relevance for mitigating climate change by means of SOC storage. There have been few comprehensive studies of MRTOC and their drivers, and these have mainly been restricted to the global scale, on which climatic drivers dominate. This study used the unique combination of regional scale cropland and grassland topsoil (0-30 cm) SOC stock data and average site-specific OC input data derived from the German Agricultural Soil Inventory to elucidate the main drivers of MRTOC . Explanatory variables related to OC input composition and other soil-forming factors were used to explain the variability in MRTOC by means of a machine-learning approach. On average, OC entering German agricultural topsoils had an MRT of 21.5±11.6 years, with grasslands (29.0±11.2 years, n=465) having significantly higher MRTOC than croplands (19.4±10.7, n=1635). This was explained by the higher proportion of root-derived OC inputs in grassland soils, which was the most important variable for explaining MRTOC variability at a regional scale. Soil properties such as clay content, soil group, C:N ratio and groundwater level were also important, indicating that MRTOC is driven by a combination of site properties and OC input composition. However, the great importance of root-derived OC inputs indicated that MRTOC can be actively managed, with maximisation of root biomass input to the soil being a straightforward means to extend the time that assimilated C remains in the soil and consequently also increase SOC stocks.

Poeplau Christopher, Don Axel, Schneider Florian


General General

Classification of masked image data.

In PloS one ; h5-index 176.0

Data classification is one of the most commonly used applications of machine learning. The are many developed algorithms that can work in various environments and for different data distributions that perform this task with excellence. Classification algorithms, just like other machine learning algorithms have one thing in common: in order to operate on data, they must see the data. In the present world, where concerns about privacy, GDPR (General Data Protection Regulation), business confidentiality and security are growing bigger and bigger; this requirement to work directly on the original data might become, in some situations, a burden. In this paper, an approach to the classification of images that cannot be directly accessed during training has been made. It has been shown that one can train a deep neural network to create such a representation of the original data that i) without additional information, the original data cannot be restored, and ii) that this representation-called a masked form-can still be used for classification purposes. Moreover, it has been shown that classification of the masked data can be done using both classical and neural network-based classifiers.

Lis Kamila, KoryciƄski Mateusz, Ciecierski Konrad A


General General

Multitask learning over shared subspaces.

In PLoS computational biology

This paper uses constructs from machine learning to define pairs of learning tasks that either shared or did not share a common subspace. Human subjects then learnt these tasks using a feedback-based approach and we hypothesised that learning would be boosted for shared subspaces. Our findings broadly supported this hypothesis with either better performance on the second task if it shared the same subspace as the first, or positive correlations over task performance for shared subspaces. These empirical findings were compared to the behaviour of a Neural Network model trained using sequential Bayesian learning and human performance was found to be consistent with a minimal capacity variant of this model. Networks with an increased representational capacity, and networks without Bayesian learning, did not show these transfer effects. We propose that the concept of shared subspaces provides a useful framework for the experimental study of human multitask and transfer learning.

Menghi Nicholas, Kacar Kemal, Penny Will


General General

A deep imputation and inference framework for estimating personalized and race-specific causal effects of genomic alterations on PSA.

In Journal of bioinformatics and computational biology

Prostate Specific Antigen (PSA) level in the serum is one of the most widely used markers in monitoring prostate cancer (PCa) progression, treatment response, and disease relapse. Although significant efforts have been taken to analyze various socioeconomic and cultural factors that contribute to the racial disparities in PCa, limited research has been performed to quantitatively understand how and to what extent molecular alterations may impact differential PSA levels present at varied tumor status between African-American and European-American men. Moreover, missing values among patients add another layer of difficulty in precisely inferring their outcomes. In light of these issues, we propose a data-driven, deep learning-based imputation and inference framework (DIIF). DIIF seamlessly encapsulates two modules: an imputation module driven by a regularized deep autoencoder for imputing critical missing information and an inference module in which two deep variational autoencoders are coupled with a graphical inference model to quantify the personalized and race-specific causal effects. Large-scale empirical studies on the independent sub-cohorts of The Cancer Genome Atlas (TCGA) PCa patients demonstrate the effectiveness of DIIF. We further found that somatic mutations in TP53, ATM, PTEN, FOXA1, and PIK3CA are statistically significant genomic factors that may explain the racial disparities in different PCa features characterized by PSA.

Chen Zhong, Cao Bo, Edwards Andrea, Deng Hongwen, Zhang Kun


Deep learning, PSA, autoencoders, causal effect inference, genomic alterations, imputation, prostate cancer, racial disparity

General General

A structured literature review on the interplay between emerging technologies and COVID-19 - insights and directions to operations fields.

In Annals of operations research

In recent years, emerging technologies have gained popularity and being implemented in different fields. Thus, critical leading-edge technologies such as artificial intelligence and other related technologies (blockchain, simulation, 3d printing, etc.) are transforming the operations and other traditional fields and proving their value in fighting against unprecedented COVID-19 pandemic outbreaks. However, due to this relation's novelty, little is known about the interplay between emerging technologies and COVID-19 and its implications to operations-related fields. In this vein, we mapped the extant literature on this integration by a structured literature review approach and found essential outcomes. In addition to the literature mapping, this paper's main contributions were identifying literature scarcity on this hot topic by operations-related fields; consequently, our paper emphasizes an urgent call to action. Also, we present a novel framework considering the primary emerging technologies and the operations processes concerning this pandemic outbreak. Also, we provided an exciting research agenda and four propositions derived from the framework, which are collated to operations processes angle. Thus, scholars and practitioners have the opportunity to adapt and advance the framework and empirically investigate and validate the propositions for this and other highly disruptive crisis.

Queiroz Maciel M, Fosso Wamba Samuel


Artificial intelligence, COVID-19, Emerging technologies, Structured literature review