Machine Learning Assisted Intraoperative Assessment of Brain Tumor Margins Using HRMAS NMR Spectroscopy

Cakmakci, Doruk; Karakaslar, Emin Onur; Ruhland, Elisa; Chenard, Marie–Pierre; Proust, F.; Piotto, Martial; Namer, Izzie Jacques; Çiçek, A. Ercüment

doi:10.1101/2020.02.24.20026955

Cited by 5 publications

(6 citation statements)

References 41 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…When the performance of baseline models is considered in this setup, RSF appears to be a particularly strong baseline with a median c-index of 67.8%. This is also expected since the power of random forest-based models on HRMAS NMR metabolomics data has been demonstrated previously on a cohort of similar size [29,30]. Still, PiDeeL achieves better performance than RSF while providing information about possibly active metabolic pathways by learning to pool information from metabolite sets encoded in its first layer.…”

Section: Survival Analysissupporting

confidence: 62%

“…Machine learning techniques that learn from the NMR signal to distinguish healthy and tumor tissue, as well as benign and aggressive tumors, have been shown to be successful. Yet, the performance was limited due to small training set sizes [29, 30] which prohibit using complex architectures like deep neural networks which learn a hierarchical composition of complex features. We, for the first time, are able to use such a hierarchically complex model for distinguishing benign and aggressive tumors.…”

Section: Discussionmentioning

confidence: 99%

“…machine learning) has the promise to be a valuable companion to the manual feedback loop depicted above due to the high-dimensional, complex and structured nature of preprocessed HRMAS NMR spectra. Indeed, various supervised machine learning models have already been explored in this context to provide automated feedback with regard to pathological classification [29,30] of tissue specimens resected during glioma surgery. The dataset used for these studies [31] consists of HRMAS NMR spectra of 565 glioma and healthy control tissue samples.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

PiDeeL: Pathway-Informed Deep Learning Model for Survival Analysis and Pathological Classification of Gliomas

Kaynar

Cakmakci

Bund

et al. 2022

Preprint

Self Cite

View full text Add to dashboard Cite

Online assessment of the tumor pathology during surgery is an important task to give feed-back to the surgeon who can be more liberal or conservative in the resection based on the input. While there are methods that perform metabolomics-based online tumor grade prediction, their performance and model complexities are limited by the small dataset sizes. Here, we propose a pathway-informed deep learning model, PiDeeL, to perform survival analysis simultaneously for a better prognostic assessment. We show that incorporating the pathway information into the model architecture and using the multitask learning framework reduce the model complexity and enable us to use deeper architectures with better performance. We show that PiDeeL improves the grade prediction performance of the state-of-the-art in terms of the Area Under the ROC Curve (AUC-ROC) by 3.38% and the Area Under the Precision-Recall Curve (AUC-PR) by 4.06% and survival analysis performance based on the time-dependent concordance index (c-index) by 2.06%. Analyzing the importance of input metabolites and neurons representing pathways provides insights into tumor metabolism. We foresee that using this model in the surgery room will help surgeons adjust the surgery plan on the fly and will result in improved prognosis due to optimized surgical procedures. The code is released athttps://github.com/ciceklab/PiDeeL. The data used in this study is released athttps://zenodo.org/record/7228791.

show abstract

Section: Survival Analysissupporting

confidence: 62%

Section: Discussionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

PiDeeL: Pathway-Informed Deep Learning Model for Survival Analysis and Pathological Classification of Gliomas

Kaynar

Cakmakci

Bund

et al. 2022

Preprint

Self Cite

View full text Add to dashboard Cite

show abstract

“…Also, machine learning can be employed to quantify the total tumor volume, the extension of the resection area, the presence of PSM (i.e., the main areas in which nuclear probes are employed), and the correlation of image features with clinical endpoints 81 . Studies have already reported the improvements obtained when using CNN for tumor margin classification in head‐and‐neck cancer, 92 brain tumors 93 or breast malignancies, 94 among others.…”

Section: The “Ideal” Probe Design Considerationsmentioning

confidence: 99%

Nuclear‐medicine probes: Where we are and where we are going

González-Montoro

Donoso²,

Konstantinou³

et al. 2022

Medical Physics

View full text Add to dashboard Cite

Nuclear medicine probes turned into the key for the identification and precise location of sentinel lymph nodes and other occult lesions (i.e., tumors) by using the systemic administration of radiotracers. Intraoperative nuclear probes are key in the surgical management of some malignancies as well as in the determination of positive surgical margins, thus reducing the extent and potential surgery morbidity. Depending on their application, nuclear probes are classified into two main categories, namely, counting and imaging. Although counting probes present a simple design, are handheld (to be moved rapidly), and provide only acoustic signals when detecting radiation, imaging probes, also known as cameras, are more hardware‐complex and also able to provide images but at the cost of an increased intervention time as displacing the camera has to be done slowly. This review article begins with an introductory section to highlight the relevance of nuclear‐based probes and their components as well as the main differences between ionization‐ (semiconductor) and scintillation‐based probes. Then, the most significant performance parameters of the probe are reviewed (i.e., sensitivity, contrast, count rate capabilities, shielding, energy, and spatial resolution), as well as the different types of probes based on the target radiation nature, namely: gamma (γ), beta (β) (positron and electron), and Cherenkov. Various available intraoperative nuclear probes are finally compared in terms of performance to discuss the state‐of‐the‐art of nuclear medicine probes. The manuscript concludes by discussing the ideal probe design and the aspects to be considered when selecting nuclear‐medicine probes.

show abstract

“…Recently, Cakmakci et al [25] demonstrated that using 1 H HRMAS NMR datasets (n=568) and machine learning based approaches, they can stratify glioma patients from controls with high accuracy and precision (mean AUC = 85.6%). Furthermore, they also showed that the models are predictive in terms of classifying the malignant and benign tumor samples.…”

Section: Hoch Et Al Uses An Iterative Approach Called Maximum Entropmentioning

confidence: 99%

Predicting Carbon Spectrum in Heteronuclear Single Quantum Coherence Spectroscopy for Online Feedback During Surgery

Karakaslar

Coskun

Outilaft

et al. 2020

IEEE/ACM Trans. Comput. Biol. and Bioinf.

Self Cite

View full text Add to dashboard Cite

Angle Spinning (HRMAS) Nuclear Magnetic Resonance (NMR) is a reliable technology used for detecting metabolites in solid tissues. Fast response time enables guiding surgeons in real time, for detecting tumor cells that are left over in the excision cavity. However, severe overlap of spectral resonances in 1D signal often render distinguishing metabolites impossible. In that case, Heteronuclear Single Quantum Coherence Spectroscopy (HSQC) NMR is applied which can distinguish metabolites by generating 2D spectra (1 H-13 C). Unfortunately, this analysis requires much longer time and prohibits real time analysis. Thus, obtaining 2D spectrum fast has major implications in medicine. In this study, we show that using multiple multivariate regression and statistical total correlation spectroscopy, we can learn the relation between the 1 H and 13 C dimensions. Learning is possible with small sample sizes and without the need for performing the HSQC analysis, we can predict the 13 C dimension by just performing 1 H HRMAS NMR experiment. We show on a rat model of central nervous system tissues (80 samples, 5 tissues) that our methods achieve 0.971 and 0.957 mean R 2 values, respectively. Our tests on 15 human brain tumor samples show that we can predict 104 groups of 39 metabolites with 97% accuracy. Finally, we show that we can predict the presence of a drug resistant tumor biomarker (creatine) despite obstructed signal in 1 H dimension. In practice, this information can provide valuable feedback to the surgeon to further resect the cavity to avoid potential recurrence.

show abstract

Machine Learning Assisted Intraoperative Assessment of Brain Tumor Margins Using HRMAS NMR Spectroscopy

Cited by 5 publications

References 41 publications

PiDeeL: Pathway-Informed Deep Learning Model for Survival Analysis and Pathological Classification of Gliomas

PiDeeL: Pathway-Informed Deep Learning Model for Survival Analysis and Pathological Classification of Gliomas

Nuclear‐medicine probes: Where we are and where we are going

Predicting Carbon Spectrum in Heteronuclear Single Quantum Coherence Spectroscopy for Online Feedback During Surgery

Contact Info

Product

Resources

About