CamDavidsonPilon/lifelines: v0.25.4

Background Analysing distributed medical data is challenging because of data sensitivity and various regulations to access and combine data. Some privacy-preserving methods are known for analyzing horizontally-partitioned data, where different organisations have similar data on disjoint sets of people. Technically more challenging is the case of vertically-partitioned data, dealing with data on overlapping sets of people. We use an emerging technology based on cryptographic techniques called secure multi-party computation (MPC), and apply it to perform privacy-preserving survival analysis on vertically-distributed data by means of the Cox proportional hazards (CPH) model. Both MPC and CPH are explained. Methods We use a Newton-Raphson solver to securely train the CPH model with MPC, jointly with all data holders, without revealing any sensitive data. In order to securely compute the log-partial likelihood in each iteration, we run into several technical challenges to preserve the efficiency and security of our solution. To tackle these technical challenges, we generalize a cryptographic protocol for securely computing the inverse of the Hessian matrix and develop a new method for securely computing exponentiations. A theoretical complexity estimate is given to get insight into the computational and communication effort that is needed. Results Our secure solution is implemented in a setting with three different machines, each presenting a different data holder, which can communicate through the internet. The MPyC platform is used for implementing this privacy-preserving solution to obtain the CPH model. We test the accuracy and computation time of our methods on three standard benchmark survival datasets. We identify future work to make our solution more efficient. Conclusions Our secure solution is comparable with the standard, non-secure solver in terms of accuracy and convergence speed. The computation time is considerably larger, although the theoretical complexity is still cubic in the number of covariates and quadratic in the number of subjects. We conclude that this is a promising way of performing parametric survival analysis on vertically-distributed medical data, while realising high level of security and privacy.

show abstract

“…The analysis was performed on the following default lifelines survival datasets [33] after filtering for missing values:…”

Section: Accuracymentioning

confidence: 99%

“…The datasets generated and/or analysed during the current study are available in the Lifelines repository [33], https:// github. com/ CamDa vidso nPilon/ lifel ines/ tree/ 97c45 5d13c f2aab a5b99 abd6b 01476 ce441 5b6d3/ lifel ines/ datas ets.…”

Section: Supplementary Informationmentioning

confidence: 99%

Accurate training of the Cox proportional hazards model on vertically-partitioned data while preserving privacy

Kamphorst

Rooijakkers

Veugen

et al. 2022

BMC Med Inform Decis Mak

View full text Add to dashboard Cite

show abstract

“…Cox Proportional Hazards modelAfter removing highly correlated variables, CPH model is developed using Python's lifeline package[165] starting with all the remained variables and iteratively removing insignificant variables. Developing a CPH model over the train dataset using 10-fold cross validation, a C-Index of 0.57 is achieved on the test set.…”

mentioning

confidence: 99%

Pedestrian Dynamics in Smart Cities: Ubiquitous Sensing, Interactions, and Models

Kalatian

2023

Preprint

View full text Add to dashboard Cite

The dissertation explores the influence of emerging technologies on the dynamics of urban areas from a pedestrian-oriented perspective. To evaluate this influence, novel data sources and modern data-driven approaches replace traditional data collection methods and modelling techniques. Part of the dissertation establishes new tools and techniques to predict and explain pedestrian behaviour that concerns the future dynamics in urban areas. Specifically, this dissertation considers: (i) what are the novel data sources that can capture detailed information on pedestrian behaviour and how can they be used efficiently? (ii) how to extract valuable information from these new high- dimensional data sources? (iii) how to utilize the data and tools to predict pedestrian behaviour in the future urban environment? and (iv) how to interpret and explain pedestrian behaviour in the context of smart cities? This thesis is based on four articles introduced in Chapters 3 to 6. Chapter 3 introduces a semi-supervised residual network for transportation mode detection using passively collected labelled and unlabelled Wi-Fi signal data. Chapter 4 provides a survival analysis to model the wait time behaviour of a crossing pedestrian and analyze the effect of smartphone distraction on pedestrian behaviour. Virtual reality is used in this chapter as a means of data collection in a controlled environment. Chapter 5 highlights pedestrian crossing behaviour in the presence of automated vehicles. A large virtual reality data collection campaign is conducted to understand pedestrian behaviour in futuristic scenarios. Data-driven survival analysis is developed to analyze pedestrian wait time before mid-block unsignalized crossings. By using a post-hoc model interpretation, the contributing factors to pedestrian wait time are assessed. In Chapter 6, a neural network architecture is developed to incorporate sequential time-series data and contextual information to predict pedestrian trajectory. The proposed framework is applied to the virtual reality dataset. Methodological and data collection frameworks explored in this dissertation provide solutions for detecting, modelling and predicting pedestrian behaviour in a futuristic context. This dissertation con- tributes to the field of transportation by proposing alternative data collection methods, developing novel data-driven methodologies and analyzing pedestrian behaviour in the context of automated vehicles.

show abstract

Glycoproteomics-based signatures for tumor subtyping and clinical outcome prediction of high-grade serous ovarian cancer

Pan

Sun

et al. 2020

Nat Commun

View full text Add to dashboard Cite

Inter-tumor heterogeneity is a result of genomic, transcriptional, translational, and post-translational molecular features. To investigate the roles of protein glycosylation in the heterogeneity of high-grade serous ovarian carcinoma (HGSC), we perform mass spectrometry-based glycoproteomic characterization of 119 TCGA HGSC tissues. Cluster analysis of intact glycoproteomic profiles delineates 3 major tumor clusters and 5 groups of intact glycopeptides. It also shows a strong relationship between N-glycan structures and tumor molecular subtypes, one example of which being the association of fucosylation with mesenchymal subtype. Further survival analysis reveals that intact glycopeptide signatures of mesenchymal subtype are associated with a poor clinical outcome of HGSC. In addition, we study the expression of mRNAs, proteins, glycosites, and intact glycopeptides, as well as the expression levels of glycosylation enzymes involved in glycoprotein biosynthesis pathways in each tumor. The results show that glycoprotein levels are mainly controlled by the expression of their individual proteins, and, furthermore, that the glycoprotein-modifying glycans correspond to the protein levels of glycosylation enzymes. The variation in glycan types further shows coordination to the tumor heterogeneity. Deeper understanding of the glycosylation process and glycosylation production in different subtypes of HGSC may provide important clues for precision medicine and tumor-targeted therapy.

show abstract

CamDavidsonPilon/lifelines: v0.25.4

Cited by 7 publications

References 0 publications

Accurate training of the Cox proportional hazards model on vertically-partitioned data while preserving privacy

Accurate training of the Cox proportional hazards model on vertically-partitioned data while preserving privacy

Pedestrian Dynamics in Smart Cities: Ubiquitous Sensing, Interactions, and Models

Glycoproteomics-based signatures for tumor subtyping and clinical outcome prediction of high-grade serous ovarian cancer

Contact Info

Product

Resources

About