Explaining Therapy Predictions with Layer-Wise Relevance Propagation in Neural Networks

Yang, Yinchong; Tresp, Volker; Wunderle, Marius; Fasching, Peter A.

doi:10.1109/ichi.2018.00025

Cited by 55 publications

(49 citation statements)

References 20 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Other studies include an evaluation on a synthetic task: Yang et al (2018) generated random sequences of MNIST digits and train an LSTM to predict if a sequence contains zero digits or not, and verify that the explanation indeed assigns a high relevance to the zero digits' positions.…”

Section: Previous Workmentioning

confidence: 99%

Evaluating Recurrent Neural Network Explanations

Arras¹,

Osman²,

Müller³

et al. 2019

Proceedings of the 2019 ACL Workshop BlackboxNLP: Analyzing and Interpreting Neural Networks for NLP

View full text Add to dashboard Cite

Recently, several methods have been proposed to explain the predictions of recurrent neural networks (RNNs), in particular of LSTMs. The goal of these methods is to understand the network's decisions by assigning to each input variable, e.g., a word, a relevance indicating to which extent it contributed to a particular prediction. In previous works, some of these methods were not yet compared to one another, or were evaluated only qualitatively. We close this gap by systematically and quantitatively comparing these methods in different settings, namely (1) a toy arithmetic task which we use as a sanity check, (2) a five-class sentiment prediction of movie reviews, and besides (3) we explore the usefulness of word relevances to build sentence-level representations. Lastly, using the method that performed best in our experiments, we show how specific linguistic phenomena such as the negation in sentiment analysis reflect in terms of relevance patterns, and how the relevance visualization can help to understand the misclassification of individual samples.

show abstract

Section: Previous Workmentioning

confidence: 99%

Evaluating Recurrent Neural Network Explanations

Arras¹,

Osman²,

Müller³

et al. 2019

Proceedings of the 2019 ACL Workshop BlackboxNLP: Analyzing and Interpreting Neural Networks for NLP

View full text Add to dashboard Cite

show abstract

“…where w + ij corresponds to the positive weights w ij and stabilizes numerical computations (Yang et al, 2018) We set to 1 −10 . Equation 7depicts the z + rule coming from deep Taylor decomposition (Montavon et al, 2017) The z + rule is commonly applied to the convolutional and fully connected layers.…”

Section: Graph Convolutional Neural Network and Layer-wise Relevance mentioning

confidence: 99%

“…Decisions proposed by neural networks have to be explained for the application in the clinical domain (Yang et al, 2018) Furthermore, the European Union's new General Data Protection Regulation (GDPR) restricted automated decision making produced by, e.g., algorithms (2018 reform of EU data protection rules 2018) Article 13 Information to be provided where personal data are collected from the data subject specifies that the data controller (e.g. clinics) should provide the data subject (e.g.…”

Section: Introductionmentioning

confidence: 99%

“…Article 22 Automated individual decision-making, including profiling states that "The data subject shall have the right not to be subject to a decision based solely on automated processing", unless the data subject gives a consent with it (paragraph 2.c). Therefore, the combination of explainability and the expressiveness of deep neural networks is yet a task to work on (Yang et al, 2018) Explanation methods for complex nonlinear models such as neural networks (including convolutional) aim to interpret classification decisions of a machine learning model in terms of input variables. These methods can be categorized into two groups (Montavon et al, 2017) functional approaches and message passing approaches.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Explaining decisions of Graph Convolutional Neural Networks: patient-specific molecular subnetworks responsible for metastasis prediction in breast cancer

Chereda

Bleckmann

Menck

et al. 2020

Preprint

View full text Add to dashboard Cite

MotivationContemporary deep learning approaches show cutting-edge performance in a variety of complex prediction tasks. Nonetheless, the application of deep learning in healthcare remains limited since deep learning methods are often considered as non-interpretable black-box models. Layer-wise Relevance Propagation (LRP) is a technique to explain decisions of deep learning methods. It is widely used to interpret Convolutional Neural Networks (CNNs) applied on image data. Recently, CNNs started to extend towards non-euclidean domains like graphs. Molecular networks are commonly represented as graphs detailing interactions between molecules. Gene expression data can be assigned to the vertices of these graphs. In other words, gene expression data can be structured by utilizing molecular network information as prior knowledge. Graph-CNNs can be applied to structured gene expression data, for example, to predict metastatic events in breast cancer. Therefore, there is a need for explanations showing which part of a molecular network is relevant for predicting an event, e.g. distant metastasis in cancer, for each individual patient.ResultsWe extended the procedure of LRP to make it available for Graph-CNN and tested its applicability on a large breast cancer dataset. We present Graph Layer-wise Relevance Propagation (GLRP) as a new method to explain the decisions made by Graph-CNNs. We demonstrate a sanity check of the developed GLRP on a hand-written digits dataset, and then applied the method on gene expression data. We show that GLRP provides patient-specific molecular subnetworks that largely agree with clinical knowledge and identify common as well as novel, and potentially druggable, drivers of tumor progression. As a result this method could be potentially highly useful on interpreting classification results on the individual patient level, as for example in precision medicine approaches or a molecular tumor board.Availabilityhttps://gitlab.gwdg.de/UKEBpublic/graph-lrphttps://frankkramer-lab.github.io/MetaRelSubNetVis/Contacttim.beissbarth@bioinf.med.uni-goettingen.de

show abstract

“…Im Rahmen des PRAEGNANT-Netzwerks [42] werden in Deutschland z. B. Methoden des Machine Learning genutzt mit dem Ziel, Therapieentscheidungen zu optimieren [43,44]. Mittels Encodings mit rekurrenten neuronalen Netzwerken und einem sogenannten Tensor Decoding konnten bereits sinnvolle Prädiktionen des optimalen klinischen Vorgehens erreicht werden [43,44].…”

Section: Unterstützung Der Klinischen Entscheidungsfindungunclassified

Translationale Highlights Mamma- und Ovarialkarzinom 2019 – Immuntherapien, DNA-Reparatur, PI3K-Inhibition und CDK4/6-Therapien

Hartkopf

Müller

Wöckel

et al. 2020

Senologie - Zeitschrift für Mammadiagnostik und -therapie

Self Cite

View full text Add to dashboard Cite

ZusammenfassungIm Rahmen von derzeit laufenden Studien werden in naher Zukunft wichtige translationale Fragestellungen von klinischer Relevanz geklärt werden. Zum einen muss die Rolle der PD‑L1-Expression, nachdem sie für die Indikationsstellung für Atezolizumab in der 1. Therapielinie bei Patientinnen mit einem metastasierten triplenegativen Mammakarzinom (TNBC) von Relevanz ist, weiter geklärt werden. In einer neoadjuvanten Studie beim TNBC mit Pembrolizumab konnte kein Zusammenhang zwischen Wirksamkeit und PD‑L1-Expression gefunden werden. Sowohl mit als auch ohne Expression war die pCR höher, wenn Pembrolizumab zur Chemotherapie hinzugefügt wurde. Des Weiteren wird für die bei Keimbahnmutation von BRCA1/2 zugelassenen PARP-Inhibitoren nach weiteren Patientengruppen gesucht, in denen eine Wirksamkeit besteht. Dieses sind z. B. Patientinnen mit Mutationen in anderen Genen, die an der homologen Rekombination beteiligt sind, oder Patientinnen mit Tumoren, die bei globalen Tests zu homologen Rekombinationsdefizienzen (HRDTests) eine Auffälligkeit zeigen. Auch die Frage, ob und mit welchen Chemotherapie-Kombinationspartnern eine PARP-Inhibition zusammen gegeben werden kann, wird derzeit sowohl beim Mamma- als auch beim Ovarialkarzinom untersucht. Während sich die Daten zum verbesserten Gesamtüberleben bei den CDK4/6-Inhibitoren konsolidieren, wächst auch das Wissen um molekulare Veränderungen unter der Therapie und bei der Progression unter der Therapie. Sowohl die Anhäufung von PI3K-Mutationen als auch PTEN-Veränderungen könnten bei der Planung von Folgetherapien eine Rolle spielen. Diese Übersichtsarbeit fasst diese aktuellen Entwicklungen beim Mammakarzinom und teilweise auch beim Ovarialkarzinom zusammen.

show abstract

Explaining Therapy Predictions with Layer-Wise Relevance Propagation in Neural Networks

Cited by 55 publications

References 20 publications

Evaluating Recurrent Neural Network Explanations

Evaluating Recurrent Neural Network Explanations

Explaining decisions of Graph Convolutional Neural Networks: patient-specific molecular subnetworks responsible for metastasis prediction in breast cancer

Translationale Highlights Mamma- und Ovarialkarzinom 2019 – Immuntherapien, DNA-Reparatur, PI3K-Inhibition und CDK4/6-Therapien

Contact Info

Product

Resources

About