Supplementary Table S8: Machine learning assisting neurosurgical care of other disease conditions

Authors / Year / Title / Journal / Stage within neurosurgical care / Disease condition / Algorithm / Input features / Application / Results/Conclusion
Hydrocephalus
Habibi, Z., Ertiaei, A., Nikdad, M. S., Mirmohseni, A. S., Afarideh, M., Heidari, V., Saberi, H., Rezaei, A. S. andNejat, F. / 2016 / Predicting ventriculoperitoneal shunt infection in children with hydrocephalus using artificial neural network / Child's Nervous System / Neurosurgical outcome prediction / Hydrocephalus / ANN / Clinical / Predict post-operative complication / Using these seven input variables, ANN and LR models predicted shunt infection with an accuracy of 83.1% (AUC; 91.98%, 95% CI) and 55.7% (AUC; 76.5, 95% CI), respectively.
Azimi, P. and Mohammadi, H. R. / 2014 / Predicting endoscopic third ventriculostomy success in childhood hydrocephalus: an artificial neural network analysis / J NeurosurgPediatr / Neurosurgical outcome prediction / Hydrocephalus / Multiple (ANN, LR) / Clinical / Predict post-operative complication / Compared with the ETVSS, CCHU ETV Success Score, and the logistic regression models, the ANN model showed better results, with an accuracy rate of 95.1%, a Hosmer-Lemeshow statistic of 41.2, and an area under the curve of 0.87.
Multiple Disease conditions
Xu, F., Zhou, W., Zhen, Y. and Yuan, Q. / 2014 / Classification of motor imagery tasks for electrocorticogram based brain-computer interface / Biomedical Engineering Letters / Neurophysiological monitoring / Multiple (TBI, Hydrocephalus) / SVM / Clinical / Non-invasive ICP assessment / The ICP estimation results on a dataset consisting of 446 entries from 23 patients show that the mean ICP error by the nonlinear approaches can be reduced to below 6.0 mmHg compared to 6.7 mmHg of the original approach. The statistical test also demonstrates that the ICP error by the proposed nonlinear kernel approaches is statistically smaller than that estimated with the original linear model (p < 0.05).
Kim, S., Hamilton, R., Pineles, S., Bergsneider, M. and Hu, X. / 2013 / Noninvasive intracranial hypertension detection utilizing semisupervised learning / IEEE Trans Biomed Eng / Neurophysiological monitoring / Multiple (TBI, Vascular, Hydrocephalus) / LDA / Ultrasound / Non-invasive ICP assessment / Our simulation results demonstrate that the predictive accuracy (area under the curve) of the semi-supervised intracranial hypertension detection method can be as high as 92% while that of the supervised intracranial hypertension detection method is only around 82%. It should be noted that the predictive accuracy of the pulsatility index (PI)-based intracranial hypertension detection method is as low as 59%.
Xie, T., Zhang, D., Wu, Z., Chen, L. and Zhu, X. / 2015 / Classifying multiple types of hand motions using electrocorticography during intraoperative awake craniotomy and seizure monitoring processes-case studies / Frontiers in Neuroscience / Neurophysiological monitoring / Multiple (Tumor, Epilepsy) / LDA / iEEG / BCI — Classification of motor imaginary tasks / Subjects performed three types of hand motions (Grasp, Thumb-finger motion and Index-finger motion) contralateral to the motor cortex covered with iEEG electrodes. The three-class accuracy were 90.17, 96.00, 91.77, and 92.95% respectively.
Mitchell, T. J., Hacker, C. D., Breshears, J. D., Szrama, N. P., Sharma, M., Bundy, D. T., Pahwa, M., Corbetta, M., Snyder, A. Z., Shimony, J. S. and Leuthardt, E. C. / 2013 / A novel data-driven approach to preoperative mapping of functional cortex using resting-state functional magnetic resonance imaging / Neurosurgery / Presurgical Planning / Multiple (Tumor, Epilepsy) / ANN / MRI / Identify eloquent area / The algorithm robustly identified all networks in all patients, including those with distorted anatomy. When all electrocortical stimulation-positive sites were considered for motor and language, rs-fMRI had AUCs of 0.80 and 0.64, respectively. When the electrocortical stimulation-positive sites were analyzed pairwise, rs-fMRI had AUCs of 0.89 and 0.76 for motor and language, respectively.
Neurosurgery in General
Wang, W. and Zhang, C. / 2010 / A brain shift correction model based on the fuzzy support vector machines with different constant term / Sheng Wu Yi Xue Gong Cheng XueZaZhi / Intraoperative Assistance / Neurosurgery in General / Multiple (Fuzzy classifier, KNN, PCA) / MRI / Align preoperative imaging with intraoperative guiding system / The results of validating the model by the leave-one-out method unveil that the approach recapitulated 90% of the shift, thus indicating that the correction model based on the fuzzy classifier with different constant term can be used to predict the brain shift with clinically acceptable accuracy.
Nucci, C. G., De Bonis, P., Mangiola, A., Santini, P., Sciandrone, M., Risi, A. and Anile, C. / 2016 / Intracranial pressure wave morphological classification: automated analysis and clinical validation / Acta Neurochir (Wien) / Neurophysiological monitoring / Neurosurgery in General / ANN / ICP sensors / Detect ICP abnormalities / The overall concordance in cerebrospinal fluid pulse pressure waveform (CSFPPW) classification between the expert examiner and the ANN was 88.3 %.
Scalzo, F., Liebeskind, D. and Hu, X. / 2013 / Reducing false intracranial pressure alarms using morphological waveform features / IEEE Trans Biomed Eng / Neurophysiological monitoring / Neurosurgery in General / Multiple (LDA, SVM) / ICP sensors / Detect ICP abnormalities / The resulting features lead to a decrease of 30% of false ICP alarms without compromising sensitivity.
Scalzo, F., Asgari, S., Kim, S., Bergsneider, M. and Hu, X. / 2012 / Bayesian tracking of intracranial pressure signal morphology / ArtifIntell Med / Neurophysiological monitoring / Neurosurgery in General / BL / ICP sensors / Detect ICP abnormalities / On artificialy distorted ICP sequences, the average error in latency in comparision with Morphological Clustering and Analysis of ICP Pulse (MOCAIP) detector was reduced as follows: 11.88-8.09 ms, 11.80-6.90 ms, and 11.76-7.46 ms for the first, second, and third peak, respectively.
Mariak, Z., Swiercz, M., Krejza, J., Lewko, J. and Lyson, T. / 2000 / Intracranial pressure processing with artificial neural networks: classification of signal properties / Acta Neurochir (Wien) / Neurophysiological monitoring / Neurosurgery in General / ANN / ICP sensors / Detect ICP abnormalities / The results were approximated to a 70% rate of judgements consistent with the expert scoring. Nevertheless, the method based on the assessment of global parameters from the ICP record looks more promising, because it leaves the possibility for modification of the set of parameters analysed.
Campillo-Gimenez, B., Garcelon, N., Jarno, P., Chapplain, J. M. and Cuggia, M. / 2013 / Full-text automated detection of surgical site infections secondary to neurosurgery in Rennes, France / Stud Health Technol Inform / Neurophysiological monitoring / Neurosurgery in General / NLP / Free-text of electronic health record / Detection of surgical site infection / The text detection was compared to the conventional strategy based on self-declaration and to the automated detection using the diagnosis-related group database. The text-mining approach showed the best detection accuracy, with recall and precision equal to 92% and 40% respectively, and confirmed the interest of reusing full-text medical reports to perform automated detection of surgical side infections
Kanas, V. G., Mporas, I., Benz, H. L., Sgarbas, K. N., Bezerianos, A. and Crone, N. E. / 2014 / Joint spatial-spectral feature space clustering for speech activity detection from ecog signals / IEEE Transactions on Biomedical Engineering / Intraoperative Assistance / Neurosurgery in General / SVM / iEEG / Identify eloquent area / We found that the optimal frequency resolution to detect speech activity from iEEG signals was 8 Hz, achieving 98.8% accuracy by employing SVMs as a classifier.
Nicolaou, N., Hourris, S., Alexandrou, P. and Georgiou, J. / 2012 / EEG-based automatic classification of 'awake' versus 'anesthetized' state in general anesthesia using Granger causality / PLoS One / Intraoperative Assistance / Neurosurgery in General / Multiple (LDA, SVM) / EEG / Monitor depth of anesthesia / Features derived from the granger causality estimates resulted in classification of 'awake' and 'anesthetized' states in 21 patients with maximum average accuracies of 0.98 and 0.95, during loss and recovery of consciousness respectively.
Seddighi, A. S., Zadeh, A. M., Seddighi, A. and Zali, A. / 2012 / Accuracy of non-invasive intracranial pressure measurement / Central European Journal of Medicine / Neurophysiological monitoring / Neurosurgery in General / Not Specified / Clinical / Non-invasive ICP assessment / In the test group, by invasive method, the mean ICP± 2SD was 17.1 ± 6.6 mmHg and using non-invasive method, the mean ICP ± 2SD was 16.5 ± 5.4 mmHg. The calculated error was 4.6 mmHg using root mean square errors. The average Pearson correlation between the estimated and real waveforms was 0.92.
Shieh, J. S., Chou, C. F., Huang, S. J. and Kao, M. C. / 2004 / Intracranial pressure model in intensive care unit using a simple recurrent neural network through time / Neurocomputing / Neurophysiological monitoring / Neurosurgery in General / ANN / Clinical / Non-invasive ICP assessment / Although the accuracy of the ICP model is still far from ideal, the methodology used non-invasive vital signs (i.e., MAP, HR, EtCO2, and rSO2) to predict an invasive, dangerous and expensive signal (i.e., ICP) has achieved this monitoring system more safely and flexibly in NICU
Hu, Y. C., Grossberg, M. and Mageras, G. / 2016 / Semiautomatic tumor segmentation with multimodal images in a conditional random field framework / Journal of Medical Imaging / Neurophysiological monitoring / Neurosurgery in General / Multiple (RF, SVM) / ICP sensors / Predict future ICP trends / The results showed that Pre-IH segments, using the optimal subset of metrics found by the differential evolution algorithm, can be differentiated from control segments at a specificity of 99% and sensitivity of 37% for these Pre-IH segments 5 min prior to the ICP elevation. While the sensitivity decreased to 21% for Pre-IH segments, 20 min prior to ICP elevation, the high specificity of 99% was retained.
Swiercz, M., Mariak, Z., Lewko, J., Chojnacki, K., Kozlowski, A. and Piekarski, P. / 1998 / Neural network technique for detecting emergency states in neurosurgical patients / Med BiolEngComput / Neurophysiological monitoring / Neurosurgery in General / ANN / ICP sensors / Predict future ICP trends / The methodology is still under development, and we believe that further ICP recordings will improve the accuracy of the ANN output.
Antoni, S. T., Rinast, J., Ma, X., Schupp, S. and Schlaefer, A. / 2016 / Online model checking for monitoring surrogate-based respiratory motion tracking in radiation therapy / Int J Comput Assist RadiolSurg / Intraoperative Assistance / Neurosurgery in General / SVM / Not Specified / Respiratory motion compensation during SRS / On patient data, online model checking detected 23-49 % of the episodes correctly, outperforming normalized least mean squares (LMS), wavelet-based multiscale autoregressio-LMS, recursive least squares and support vector regression by up to 544, 491, 408 and 258 %, respectively. On selected episodes, online model checking detected up to 94 % of all events.
Amiri, S., Movahedi, M. M., Kazemi, K. and Parsaei, H. / 2016 / 3D cerebral MR image segmentation using multiple-classifier system / Med BiolEngComput / Presurgical Planning / Neurosurgery in General / ANN / MRI / Segmentation of critical/target brain structure / As accurately segmenting a MR image is of paramount importance for successfully promoting the clinical application of MR image segmentation techniques, the improvement obtained by using multiple-classifier-based system is encouraging.
Liu, Y. and Dawant, B. M. / 2015 / Automatic localization of the anterior commissure, posterior commissure, and midsagittal plane in MRI scans using regression forests / IEEE J Biomed Health Inform / Presurgical Planning / Neurosurgery in General / RF / MRI / Segmentation of critical/target brain structure / Our method results in an overall error of 0.55 +/-0.30 mm for anterior commissure, 0.56 +/-0.28 mm for posterior commissure, 1.08( degrees ) +/-0.66 in the plane's normal direction, and 1.22 +/-0.73 voxels in average distance for midsagittal plane; it performs significantly better than four registration algorithms and the model-based method for anterior commissure and posterior commissure and the global symmetry-based method for midsagittal plane; We also evaluate the sensitivity of our method to image quality and parameter values. We show that it is robust to asymmetry, noise, and rotation. Computation time is 25 s.
Goncalves, N., Nikkila, J. and Vigario, R. / 2014 / Self-supervised MRI tissue segmentation by discriminative clustering / Int J Neural Syst / Presurgical Planning / Neurosurgery in General / QDA / MRI / Segmentation of critical/target brain structure / This method avoids the use of prior information, which makes it very versatile, and able to cope with different tissue types. It also returns tissue probabilities for each voxel, crucial for a good characterization of the evolution of brain lesions.
Liu, Y. and Dawant, B. M. / 2014 / Automatic detection of the anterior and posterior commissures on MRI scans using regression forests / Conf Proc IEEE Eng Med BiolSoc / Presurgical Planning / Neurosurgery in General / RF / MRI / Segmentation of critical/target brain structure / Our method results in an overall error of 0.84+/-0.41mm for anterior commissure 0.83+/-0.36mm for posterior commissure and a maximum error of 2.04mm; it performs significantly better than the model-based anterior/posterior commissure detection method we compare it to and better than three of the nonrigid registration methods. It is much faster than nonrigid registration methods.
Yazdani, S., Yusof, R., Riazi, A. and Karimian, A. / 2014 / Magnetic resonance image tissue classification using an automatic method / DiagnPathol / Presurgical Planning / Neurosurgery in General / SVM / MRI / Segmentation of critical/target brain structure / Our method has been validated on real images and simulated data, with desirable performance in the presence of noise and intensity inhomogeneities.
Lin, G. C., Wang, W. J., Wang, C. M. and Sun, S. Y. / 2010 / Automated classification of multi-spectral MR images using Linear Discriminant Analysis / Comput Med Imaging Graph / Presurgical Planning / Neurosurgery in General / Multiple (FCM, LDA) / MRI / Segmentation of critical/target brain structure / Experiment results reveal that unsupervised LDA segments a multi-spectral MR image much more effectively than either FMRIB's Automated Segmentation Tool (FAST) or Fuzzy C-means (FC).
Seghier, M. L., Ramlackhansingh, A., Crinion, J., Leff, A. P. and Price, C. J. / 2008 / Lesion identification using unified segmentation-normalisation models and fuzzy clustering / Neuroimage / Presurgical Planning / Neurosurgery in General / Not Specified / MRI / Segmentation of critical/target brain structure / Our analyses show a high sensitivity for detecting and delineating brain lesions with different sizes, locations, and textures.
Sun, H., Lunn, K. E., Farid, H., Wu, Z., Roberts, D. W., Hartov, A. and Paulsen, K. D. / 2005 / Stereopsis-guided brain shift compensation / IEEE transactions on medical imaging / Intraoperative Assistance / Neurosurgery in General / ANN / MRI / Segmentation of critical/target brain structure / We have found that the stereo vision system is accurate to within approximately 1 mm. Based on data from two representative clinical cases, we show that stereopsis guidance improves the accuracy of brain shift compensation both at and below the cortical surface.

Abbreviations: ANN: Artificial Neural Networks; BCI: Brain; Computer Interface; BL: Bayesion Learning; CT: Computer Tomography; DBS: Deep Brain Stimulation; DT: Decision Tree; EEG: Electro-EncephaloGraphy; FCM: Fuzzy C-Means; FIS: Fuzzy Inference System; GA: Genetic Algorithm; GB: Gradient Boosting; GMM: Gaussian Mixture Models; ICP: Intracranial Pressue; iEEG: Intracranial Electro-EncephaloGraphy; KNN: K-Nearest Neighbors; LDA: Linear Discriminant Analysis; LR: Logistic Regression; MERs: Micro-Electrode Recordings; MRI: Magnetic Resonance Imaging; NLP: Natural language processing; OLS: Ordinary Least Squares; PCA: Principal Component Analysis; PET: Positrion Emission Tomography; QDA: Quadratic Discriminant Analysis