Machine Learning in Neuro-Oncology: Can Data Analysis From 5346 Patients Change Decision-Making Paradigms?

doi:10.1016/j.wneu.2019.01.046

World Neurosurgery

Volume 124, April 2019, Pages 287-294

https://doi.org/10.1016/j.wneu.2019.01.046 Get rights and content

Background

Machine learning (ML) is an application of artificial intelligence (AI) that gives computer systems the ability to learn data, without being explicitly programmed. Currently, ML has been successfully used for optical character recognition, spam filtering, and face recognition. The aim of the present study was to review the current applications of ML in the field of neuro-oncology.

Methods

We conducted a systematic literature review using the PubMed and Cochrane databases using a keyword search for January 30, 2000 to March 31, 2018. The data were clustered for neuro-oncology scope of ML into 3 categories: patient outcome predictors, imaging analysis, and gene expression.

Results

Data from 5346 patients in 29 studies were used to develop ML-based algorithms (MLBAs) in neuro-oncology. MLBAs were used to predict the outcomes for 2483 patients, with a sensitivity range of 78%–98% and specificity range of 76%–95%. In all studies, the MLBAs had greater accuracy than the conventional ones. MLBAs for image analysis showed accuracy in diagnosing low-grade versus high-grade gliomas, ranging from 80% to 93% and 90% for diagnosing high-grade glioma versus lymphoma. Seven studies used MLBAs to analyze gene expression in neuro-oncology.

Conclusions

MLBAs in neuro-oncology have been shown to predict patients' outcomes more accurately than conventional parameters in a retrospective analysis. If their high diagnostic accuracy in imaging analysis and detection of somatic mutations are corroborated in prospective studies, the use of tissue diagnosis or liquid biopsy might be curtailed. Finally, MLBAs are promising to help guide targeted therapy, can lead to personalized medicine, and open areas of study in the cancer cellular signaling system, not otherwise known.

Introduction

Machine learning (ML) is a current application of artificial intelligence based on a field of computer science that gives computer systems the ability to learn from “big data,” without being explicitly programmed. This emerging data analysis method has been successfully used for optical character recognition, spam filtering, and face recognition. For example, support vector machines (SVMs) are learning algorithms used to perform analysis on a set of data and then provide classification. After inputting a data set of various training points with an example of each classification, the SVM ML-based algorithm (MLBA) can then group any new data provided into the various categories. Its application could be very useful in the field of medicine. The major benefit of ML is the propensity to analyze a large amount of nonlinear data to then allow for predictions to help guide medical decision-making. Thus, a shift is occurring from conventional statistical methods to ML to allow for more meaningful analysis of a greater amount of medical data, the so-called big data.

The use of ML to help guide neurosurgical care in areas such as preoperative planning, intraoperative guidance, neurophysiological monitoring, and neurosurgical outcome prediction has greatly increased in the past 5 years.¹ The main concept in MLBAs is to use a sample data set and construct a mathematical model to allow for predictions without explicit programming. MLBAs can be divided into 2 broad categories: supervised learning or unsupervised learning. With supervised learning, the sample data set will incorporate both the input data and the desired output. With unsupervised learning, the data set only has the input without the output. Although detailed concepts of ML have been previously elucidated, the goal of the present study was to review specifically the application and use of ML in the field of neuro-oncology.

Section snippets

Search Strategy and Data Extraction

The present systematic review and meta-analysis complied with the PRISMA-IPD (Preferred Reporting Items for Systematic Review and Meta-Analyses of Individual Participant Data) guidelines.² The PubMed database was searched for English-language studies reported from January 30, 2000 to March 31, 2018. The MeSH terms used were “neurosurgery,” “machine learning,” “glioma,” “spine,” “prediction,” “support vector machine,” “Bayesian network,” “decision tree,” “data mining,” and “neural network,”

Results

The 29 included studies had used data from 5346 patients to develop MLBAs in neuro-oncology (Figure 2A). These 29 studies implemented ML techniques in neuro-oncology. The increase in the number of studies and patients was significant between 2015 and 2016 (P < 0.05), with a subsequent plateau (Figure 2B). The most commonly used MLBA was SVM (Table 1).

Discussion

ML has the theoretical advantage of providing an accurate and fast interpretation of complex data and overcoming possible human error and/or bias. Our review has shown that its application in neuro-oncology has 3 major categories: predicting patient outcomes, analyzing imaging findings, and predicting gene expression. The current limitations of MLBAs have been reviewed previously.³²

Our review yielded 12 studies that had used MLBAs to predict patient outcomes, with 9 studies investigating

Conclusions

ML has the theoretical advantage to provide accurate and fast interpretation of complex data, overcoming possible human error and/or bias. With the advances in the field of ML from handcrafted features to automatic feature engineering, the accuracy of the models will certainly increase. MLBAs in neuro-oncology have shown to predict patient outcomes more accurately than conventional parameters on retrospective analysis. Additionally, if their high diagnostic accuracy in imaging analysis and

References (32)

K. Malhotra et al.
Constraint based temporal event sequence mining for glioblastoma survival prediction
J Biomed Inform
(2016)
F. Kellner-Weldon et al.
Comparison of perioperative automated versus manual two-dimensional tumor analysis in glioblastoma patients
Eur J Radiol
(2017)
Y. Li et al.
MRI features predict p53 status in lower-grade gliomas via a machine-learning approach
NeuroImage Clin
(2018)
E. Celtikci
A systematic review on machine learning in neurosurgery: the future of decision-making in patient care
Turk Neurosurg
(2018)
D. Moher et al.
Preferred reporting items for systematic reviews and meta-analyses: the PRISMA statement
BMJ
(2009)
K.E. Emblem et al.
A generic support vector machine model for preoperative glioma survival associations
Radiology
(2015)
L. Macyszyn et al.
Imaging patterns predict patient survival and molecular subtype in glioblastoma via machine learning techniques
Neuro Oncol
(2016)
E.I. Zacharaki et al.
Survival analysis of patients with high-grade gliomas based on data mining of imaging variables
AJNR Am J Neuroradiol
(2012)
J. Lao et al.
A deep learning-based radiomics model for prediction of survival in glioblastoma multiforme
Sci Rep
(2017)
Q. Li et al.
A fully-automatic multiparametric radiomics model: towards reproducible and prognostic imaging signature for prediction of overall survival in glioblastoma multiforme
Sci Rep
(2017)

L. Liu et al.

Outcome prediction for patient with high-grade gliomas from brain functional and structural networks

Med Image Comput Comput Assist Interv

(2016)

D. Nie et al.

3D deep learning for multi-modal imaging-guided survival time prediction of brain tumor patients

Med Image Comput Comput Assist Interv

(2016)

L. Papp et al.

Glioma survival prediction with the combined analysis of in vivo 11C-MET-PET, ex vivo and patient features by supervised machine learning

J Nucl Med

(2018)

A. Celejewska et al.

Stereotactic radiotherapy in epithelial ovarian cancer brain metastases patients

J Ovarian Res

(2014)

M.A. Knoll et al.

Survival of patients with multiple intracranial metastases treated with stereotactic radiosurgery: does the number of tumors matter?

Am J Clin Oncol

(2018)

S. Lee et al.

Analysis of risk factors to predict communicating hydrocephalus following gamma knife radiosurgery for intracranial schwannoma

Cancer Med

(2016)

Cited by (32)

Exploring the Landscape of Machine Learning Applications in Neurosurgery: A Bibliometric Analysis and Narrative Review of Trends and Future Directions
2024, World Neurosurgery
The field of neurosurgery has consistently represented an area of innovation and integration of technology since its inception. As such, machine learning (ML) has found its way into applications within neurosurgery relatively rapidly. Through this bibliometric review and cluster analysis, we seek to identify trends and emerging applications of ML within neurosurgery.
A bibliometric analysis was carried out in the Web of Science database on publications from January 2000 to March 2023. The full data set of the 200 most cited publications including title, author information, journal, citation count, keywords, and abstracts for each publication was evaluated in CiteSpace. CiteSpace was used to elucidate publication characteristics, trends, and topic clusters via collaborate network analysis using the Kamada-Kawai algorithm.
The 25 most cited titles were included in our analysis. Harvard University and its affiliates represented the top institution, contributing nearly 25% of publications in the literature. WORLD NEUROSURGERY was the journal with the highest net citation count of 747 (29%). Collaborative network analysis generated 12 unique clusters, the largest of which was machine learning, followed by feature importance and deep brain stimulation.
This review highlights the most impactful articles pertaining to ML in the field of neurosurgery. ML has been applied into several sub-specialties within neurosurgery to optimize patient care, with special attention to outcome predictors, patient selection, and surgical decision making.
Fundamental principles of machine learning for neurologists
2023, Pratique Neurologique - FMC
Bien que théorisées et développées depuis le XX^e siècle, ce n’est qu’en 2010 que l’on a vu une véritable démocratisation des logiciels d’apprentissage automatique, notamment dans le domaine de la santé. L’intelligence artificielle (IA) est un terme général qui englobe toutes les techniques utilisées pour effectuer un raisonnement ou des tâches traditionnellement réservées aux êtres humains. Les tâches permises par l’IA sont très variées, telles que la classification, la détection, la génération de contenu à partir de bases de données, etc., reviennent en pratique dans tous les cas à des tâches de classification automatique. L’apprentissage automatique est un sous-domaine de l’IA qui fait référence aux programmes qui calculent des règles de décision à partir de données empiriques annotées par des experts et qui optimisent (selon des indicateurs choisis par les concepteurs aussi appelé fonction de coût) la performance en termes de prédiction de résultats confrontés aux annotations des experts, par des essais itératifs prenant en compte les essais déjà réalisés. Ces programmes requièrent donc des données annotées ou classifiées (par exemple des images avec des annotations, comme la présence/absence d’une tumeur) et ont révélé leur potentiel avec des données complexes (dont le nombre de paramètres peut dépasser les capacités de l’esprit humain et/ou non perceptibles par l’humain) et structurées telles que des images, des vidéos, du texte, de la parole, etc. Parmi les méthodes les plus populaires, on peut citer les réseaux neuronaux, les « Support Vector Machines » ou les méthodes d’ensemble (bagging, boosting, random forests).
Although machine learning has been theorized and developed since the 20th century, it has only been since 2010 that we have seen a real democratization of machine learning software, particularly in the field of health. Artificial intelligence (AI) is a general term that encompasses all the techniques used to perform reasoning or tasks traditionally reserved for human beings. AI can enable a wide range of tasks, for example classification, detection, generation of content from databases, etc. In practice, everything comes down to automatic classification tasks. Machine learning is a sub-field of AI that refers to programs that calculate decision rules from empirical data annotated by experts and optimize (according to indicators chosen by the designers, also known as the cost function) performance in terms of predicting results based on the experts’ annotations, by iterative testing taking into account tests carried out previously. These programs therefore require annotated or classified data (e.g. images with annotations, such as the presence/absence of a tumor) and have revealed their potential with complex data (where the number of parameters may exceed the capabilities of the human mind and/or may not be perceptible by humans) and structured data such as images, videos, text, speech, etc. The most popular methods include neural networks, Support Vector Machines and ensemble methods (bagging, boosting, random forests).
Overall survival prediction for high-grade glioma patients using mathematical modeling of tumor cell infiltration
2023, Physica Medica
This study aimed at applying a mathematical framework for the prediction of high-grade gliomas (HGGs) cell invasion into normal tissues for guiding the clinical target delineation, and at investigating the possibility of using tumor infiltration maps for patient overall survival (OS) prediction.
A model describing tumor infiltration into normal tissue was applied to 93 HGG cases. Tumor infiltration maps and corresponding isocontours with different cell densities were produced. ROC curves were used to seek correlations between the patient OS and the volume encompassed by a particular isocontour. Area-Under-the-Curve (AUC) values were used to determine the isocontour having the highest predictive ability. The optimal cut-off volume, having the highest sensitivity and specificity, for each isocontour was used to divide the patients in two groups for a Kaplan-Meier survival analysis.
The highest AUC value was obtained for the isocontour of cell densities 1000 cells/mm³ and 2000 cells/mm³, equal to 0.77 (p < 0.05). Correlation with the GTV yielded an AUC of 0.73 (p < 0.05). The Kaplan-Meier survival analysis using the 1000 cells/mm³ isocontour and the ROC optimal cut-off volume for patient group selection rendered a hazard ratio (HR) of 2.7 (p < 0.05), while the GTV rendered a HR = 1.6 (p < 0.05).
The simulated tumor cell invasion is a stronger predictor of overall survival than the segmented GTV, indicating the importance of using mathematical models for cell invasion to assist in the definition of the target for HGG patients.
Enhancing sepsis management through machine learning techniques: A review
2022, Medicina Intensiva
Citation Excerpt :
There are applications of ML to almost all medical fields. Some recent reviews of the use of ML in several medical areas have been published, including medical image analysis,32 cardiovascular medicine,33 in critical care,34 or neuro oncology.35 The main objective of this work is to conduct a narrative review to provide an overview of how specific ML techniques can be used to improve sepsis management, focusing on the following specific tasks: detection, prediction of sepsis/shock, mortality prediction, hospital stay, costs and adherence to guides.
Sepsis is a major public health problem and a leading cause of death in the world, where delay in the beginning of treatment, along with clinical guidelines non-adherence have been proved to be associated with higher mortality. Machine Learning is increasingly being adopted in developing innovative Clinical Decision Support Systems in many areas of medicine, showing a great potential for automatic prediction of diverse patient conditions, as well as assistance in clinical decision making. In this context, this work conducts a narrative review to provide an overview of how specific Machine Learning techniques can be used to improve sepsis management, discussing the main tasks addressed, the most popular methods and techniques, as well as the obtained results, in terms of both intelligent system accuracy and clinical outcomes improvement.
La sepsis representa un problema de salud pública de primer orden y es una de las principales causas de muerte a nivel mundial. El retraso en el inicio del tratamiento, junto con la no adherencia a las guías de práctica clínica se asocian a una mayor mortalidad. El aprendizaje automático o machine learning están siendo empleados en el desarrollo de sistemas de apoyo a la decisión clínica, innovadores en muchas áreas de la medicina, mostrando un gran potencial para la predicción de diversas condiciones del paciente, así como en la asistencia durante el proceso de toma de decisiones médicas. En este sentido, este trabajo lleva a cabo una revisión narrativa para proporcionar una visión general de cómo las técnicas de machine learning pueden ser empleadas para mejorar el manejo de la sepsis, discutiendo las principales tareas que tratan de resolver, los métodos y las técnicas más empleados, así como los resultados obtenidos, tanto en términos de precisión de los sistemas inteligentes, como en la mejora de los resultados clínicos.
Machine learning workflow for the oil uptake prediction of rice flour in a batter-coated fried system
2021, Innovative Food Science and Emerging Technologies
Citation Excerpt :
Therefore, more thorough analysis incorporating multiple physicochemical features rather than focusing on a single variable is necessary to comprehensively understand the oil uptake of fried foods. Artificial intelligence (AI) indicates human intelligence exhibited by machines (Bini, 2018) and machine learning is a subset of AI that can make systems have the ability to automatically learn and improve from data without being explicitly programmed (Sarkiss & Germano, 2019). It is no doubt that machine learning as the most innovative technology is being exploited in a wide variety of fields, rapidly transforming industries and societies across the world.
Machine learning algorithms were applied to predict the oil uptake of rice flour in batter-coated fried foods, depending on its physicochemical features before frying (amylose content, pasting parameters, and batter viscosity). Low coefficients of Pearson correlation (0.29–0.61) were observed between the oil uptake of rice frying batters and other physicochemical parameters. Based on an exhaustive search method by the regsubsets function, the four experimental features (amylose content, peak time, peak temperature, and final viscosity) were selected as the best subset to affect the oil uptake of rice batters after frying, and then subjected to two machine learning algorithms – multivariable linear regression and multilayer perceptron neural network. Based on K-fold cross-validation, the experimental results were divided into 5 datasets consisting of each 80% training and 20% testing dataset. Compared to the multivariable linear regression (R² = 0.6204–0.7219), the iterative application of the multilayer perceptron model made a relatively higher prediction (R² = 0.7388–0.7781) of the oil uptake of rice flour frying batter. Thus, the multilayer perception model with a hidden layer outperformed the multivariable linear regression by showing higher R² and lower relative error.
Noninterpretive Uses of Artificial Intelligence in Radiology
2021, Academic Radiology
Citation Excerpt :
Deep learning has also been used in attempts to identify MR features that could help classify molecular subtypes of [89,90] and estimate patient prognosis in [91] patients with intracranial gliomas. A systematic review in neuro-oncology found that AI models could predict patient outcomes such as survival with higher accuracy than conventional staging and clinical risk parameters [92]. For instance, AI can predict prognosis in patients with high-grade glioma with accuracies of 91% [93].
We deem a computer to exhibit artificial intelligence (AI) when it performs a task that would normally require intelligent action by a human. Much of the recent excitement about AI in the medical literature has revolved around the ability of AI models to recognize anatomy and detect pathology on medical images, sometimes at the level of expert physicians. However, AI can also be used to solve a wide range of noninterpretive problems that are relevant to radiologists and their patients. This review summarizes some of the newer noninterpretive uses of AI in radiology.

View all citing articles on Scopus

: Conflict of interest statement: The authors declare that the article content was composed in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

: Christopher A. Sarkiss and Isabelle M. Germano are co–first authors.

View full text

Literature ReviewMachine Learning in Neuro-Oncology: Can Data Analysis From 5346 Patients Change Decision-Making Paradigms?

Background

Methods

Results

Conclusions

Introduction

Section snippets

Search Strategy and Data Extraction

Results

Discussion

Conclusions

J Biomed Inform

Eur J Radiol

NeuroImage Clin

A systematic review on machine learning in neurosurgery: the future of decision-making in patient care

Turk Neurosurg

Preferred reporting items for systematic reviews and meta-analyses: the PRISMA statement

BMJ

A generic support vector machine model for preoperative glioma survival associations

Radiology

Imaging patterns predict patient survival and molecular subtype in glioblastoma via machine learning techniques

Neuro Oncol

Survival analysis of patients with high-grade gliomas based on data mining of imaging variables

AJNR Am J Neuroradiol

A deep learning-based radiomics model for prediction of survival in glioblastoma multiforme

Sci Rep

A fully-automatic multiparametric radiomics model: towards reproducible and prognostic imaging signature for prediction of overall survival in glioblastoma multiforme

Sci Rep

Outcome prediction for patient with high-grade gliomas from brain functional and structural networks

Med Image Comput Comput Assist Interv

3D deep learning for multi-modal imaging-guided survival time prediction of brain tumor patients

Med Image Comput Comput Assist Interv

Glioma survival prediction with the combined analysis of in vivo 11C-MET-PET, ex vivo and patient features by supervised machine learning

J Nucl Med

Stereotactic radiotherapy in epithelial ovarian cancer brain metastases patients

J Ovarian Res

Survival of patients with multiple intracranial metastases treated with stereotactic radiosurgery: does the number of tumors matter?

Am J Clin Oncol

Analysis of risk factors to predict communicating hydrocephalus following gamma knife radiosurgery for intracranial schwannoma

Cancer Med

Literature Review
Machine Learning in Neuro-Oncology: Can Data Analysis From 5346 Patients Change Decision-Making Paradigms?