MR Spektroskopi Sinyalleri Kullanılarak LSTM Derin Sinir Ağları ile Beyinde Sahte Tümörlerin Tespiti

Year 2020, Ejosat Special Issue 2020 (HORA), 426 - 433, 15.08.2020

Emre Dandıl , Semih Karaca

https://doi.org/10.31590/ejosat.780675

Cited By: 2

Abstract

Manyetik rezonans spektroskopi (MRS) günümüzde beyin tümörlerinin tespitinde kullanılan müdahalesiz araçlardan biridir. Biyopsi gibi ameliyata bağlı enfeksiyon ve ölüm riski getirmediği için hekimler tarafından yaygın olarak tercih edilmektedir. MRS beyinle ilgili metabolik bir profil sunmaktadır. Tümör ve sahte tümörlerin MRS örüntüleri birbirleri ile benzerlik gösterebilmektedir. Bu sebepten dolayı beyin tümörünün doğru teşhisi ve sınıflandırılması, hastanın tedavi planlaması açısından hayati bir önem taşımaktadır. Bu çalışmada, MRS verileri kullanılarak, derin sinir ağları ile gerçek ve sahte beyin tümörlerinin sınıflandırılması sağlanmıştır. Çalışma kapsamında yürütülen deneysel çalışmalarda, LSTM (Long Short Term Memory – Uzun Kısa Süreli Bellek) ve Bi-LSTM (Bi-directional Long Short Term Memory – Çift Yönlü Uzun Kısa Süreli Bellek) derin sinir ağları mimarileri kullanılmıştır. Çalışmada INTERPRET veritabanında bulunan tümör ve sahte tümörlere ait MRS sinyal örüntüleri kullanılmıştır. LSTM sinir ağlarının eğitimi ve test edilmesi için çok sayıda tümör ve sahte tümöre ait MRS verisini elde etmek gerçek dünyada zor bir prosedürel süreç olduğundan, ağ eğitilmeden ve test edilmeden önce, MRS veriseti için veri büyütme (çoğaltma) yöntemleri ile veri sayısı çoğaltılmıştır. LSTM sinir ağları, hem veri çoğaltma olmadan hem de veri çoğaltma ile eğitilmiş ve test edilmiştir. Kullanılan LSTM sinir ağlarının eğitim ve testleri esnasında her model için tekrarlı K-kat çapraz doğrulama yöntemi kullanılmıştır. Eğitimler, her model için 5 kat ve 10 tekrar ile yapılmıştır. MRS verilerini bilgisayar destekli sınıflandırmaya dayalı bir yöntem ile sınıflandıran bu çalışma sonucunda, geliştirilen uygulama ile veri çoğaltma olmadan yapılan testlerde, kullanılan iki mimari için ortalama %81.15 doğruluk başarımı elde edilirken; veri çoğaltma yapıldıktan sonra yapılan testlerde, her iki mimari için ortalama %95.15 doğruluk başarımı elde edilmiştir.

Keywords

Beyin tümörleri, MRS, Sınıflandırma, Derin Öğrenme, LSTM, Bi-LSTM

Thanks

Bu çalışmada, sınıflandırılan MRS verilerinin yer aldığı INTERPRET veri tabanının oluşturulmasında emeği geçen tüm bilim adamlarına ve projeyi finanse eden Avrupa komisyonuna teşekkürlerimizi sunarız.

Detection of Pseudo Brain Tumors via LSTM Neural Networks using MR Spectroscopy Signals

Abstract

Magnetic resonance spectroscopy (MRS) is one of the non-invasive tools used in the detection of brain tumors today. It is widely preferred by physicians because it does not pose a risk of surgical infection and death such as biopsy. MRS provides a metabolic profile about the brain. MRS patterns of tumors and pseudo tumors can be similar to each other. For this reason, accurate diagnosis and classification of the brain tumor is vital for the treatment of the patient. In this study, the classification of real and pseudo brain tumors with deep neural networks was provided by using MRS data. In experimental studies carried out within the scope of the study, LSTM (Long Short Term Memory) and Bi-LSTM (Bidirectional Long Short Term Memory) deep neural network architectures were used. In the study, MRS signal patterns of tumors and pseudo tumors in the INTERPRET database were used. Since obtaining MRS data from a large number of tumors and pseudo tumors for the training and testing of LSTM neural networks is a difficult procedural process in the real world, the number of data has been increased by MRS data augmentation (replication) methods before the network is trained and tested. LSTM neural networks are trained and tested both with and without data augmentation methods. During the training and testing of the LSTM neural networks, repeated K-fold cross-validation method was used for each model. Neural network trainings were carried out with 5 folds and 10 repetitions for each model. As a result of this study which classifies MRS data with a method based on computer-aided classification; in the tests carried out without data augmentation with the developed application, an average of 81.15% accuracy was achieved for the 2 neural network architectures while in the tests performed after data augmentation, an average of 95.15% accuracy performance was achieved for both networks.

Keywords

Brain Tumors, MRS, Classification, Deep Learning, LSTM, Bi-LSTM

References

• Arizmendi, C., Sierra, D. A., Vellido, A., & Romero, E. (2014). Automated classification of brain tumours from short echo time in vivo MRS data using Gaussian Decomposition and Bayesian Neural Networks. Expert systems with applications, 41(11), 5296-5307.
• Arús, C., Celda, B., Dasmahaptra, S., Dupplaw, D., Gonzalez-Velez, H., Van Huffel, S., . . . Peet, A. (2006). On the design of a web-based decision support system for brain tumour diagnosis using distributed agents. Paper presented at the 2006 IEEE/WIC/ACM International Conference on Web Intelligence and Intelligent Agent Technology Workshops.
• Butzen, J., Prost, R., Chetty, V., Donahue, K., Neppl, R., Bowen, W., . . . Kim, T. (2000). Discrimination between neoplastic and nonneoplastic brain lesions by use of proton MR spectroscopy: the limits of accuracy with a logistic regression model. American journal of neuroradiology, 21(7), 1213-1219.
• Callot, V., Galanaud, D., Le Fur, Y., Confort-Gouny, S., Ranjeva, J.-P., & Cozzone, P. J. (2008). 1H MR spectroscopy of human brain tumours: a practical approach. European journal of radiology, 67(2), 268-274.
• Crain, I. D., Elias, P. S., Chapple, K., Scheck, A. C., Karis, J. P., & Preul, M. C. (2017). Improving the utility of 1 H-MRS for the differentiation of glioma recurrence from radiation necrosis. Journal of neuro-oncology, 133(1), 97-105.
• Devos, A., Lukas, L., Suykens, J., Vanhamme, L., Tate, A., Howe, F., . . . Arus, C. (2004). Classification of brain tumours using short echo time 1H MR spectra. Journal of Magnetic Resonance, 170(1), 164-175.
• Faria, A. V., Macedo Jr, F., Marsaioli, A., Ferreira, M., & Cendes, F. (2011). Classification of brain tumor extracts by high resolution ¹H MRS using partial least squares discriminant analysis. Brazilian Journal of Medical and Biological Research, 44(2), 149-164.
• Fitzmaurice, C., Allen, C., Barber, R. M., Barregard, L., Bhutta, Z. A., Brenner, H., . . . Dandona, L. (2017). Global, regional, and national cancer incidence, mortality, years of life lost, years lived with disability, and disability-adjusted life-years for 32 cancer groups, 1990 to 2015: a systematic analysis for the global burden of disease study. JAMA oncology, 3(4), 524-548.
• Georgiadis, P., Kostopoulos, S., Cavouras, D., Glotsos, D., Kalatzis, I., Sifaki, K., . . . Nikiforidis, G. (2011). Quantitative combination of volumetric MR imaging and MR spectroscopy data for the discrimination of meningiomas from metastatic brain tumors by means of pattern recognition. Magnetic resonance imaging, 29(4), 525-535.
• Graves, A., Mohamed, A.-r., & Hinton, G. (2013). Speech recognition with deep recurrent neural networks. Paper presented at the 2013 IEEE international conference on acoustics, speech and signal processing.
• Hekmatnia, A., Sabouri, M., Ghazavi, A. H., Far, P. S., Hekmatnia, F., Sofi, G. J., . . . Salehi, M. (2019). Diagnostic value of Magnetic Resonance Spectroscopy (MRS) for detection of Brain Tumors in patients. Medical Science, 23(100), 939-945.
• Hochreiter, S., & Schmidhuber, J. (1997). Long short-term memory. Neural computation, 9(8), 1735-1780.
• Hopfield, J. J. (1982). Neural networks and physical systems with emergent collective computational abilities. Proceedings of the national academy of sciences, 79(8), 2554-2558.
• Horská, A., & Barker, P. B. (2010). Imaging of brain tumors: MR spectroscopy and metabolic imaging. Neuroimaging Clinics, 20(3), 293-310.
• Hourani, R., Brant, L., Rizk, T., Weingart, J. D., Barker, P. B., & Horská, A. (2008). Can proton MR spectroscopic and perfusion imaging differentiate between neoplastic and nonneoplastic brain lesions in adults? American journal of neuroradiology, 29(2), 366-372.
• Howe, F. A., & Opstad, K. S. (2003). 1H MR spectroscopy of brain tumours and masses. NMR in Biomedicine: An International Journal Devoted to the Development and Application of Magnetic Resonance In Vivo, 16(3), 123-131.
• INTERPRET. (2002). International network for pattern recognition of tumours using magnetic resonance. .
• Kimura, T., Sako, K., Gotoh, T., Tanaka, K., & Tanaka, T. (2001). In vivo single‐voxel proton MR spectroscopy in brain lesions with ring‐like enhancement. NMR in Biomedicine: An International Journal Devoted to the Development and Application of Magnetic Resonance In Vivo, 14(6), 339-349.
• Ladd, M. E., Bachert, P., Meyerspeer, M., Moser, E., Nagel, A. M., Norris, D. G., . . . Zaiss, M. (2018). Pros and cons of ultra-high-field MRI/MRS for human application. Progress in nuclear magnetic resonance spectroscopy, 109, 1-50.
• Louis, D. N., Perry, A., Reifenberger, G., Von Deimling, A., Figarella-Branger, D., Cavenee, W. K., . . . Ellison, D. W. (2016). The 2016 World Health Organization classification of tumors of the central nervous system: a summary. Acta neuropathologica, 131(6), 803-820.
• Majos, C., Aguilera, C., Alonso, J., Julia-Sape, M., Castaner, S., Sanchez, J., . . . Arus, C. (2009). Proton MR spectroscopy improves discrimination between tumor and pseudotumoral lesion in solid brain masses. American journal of neuroradiology, 30(3), 544-551.
• Majós, C., Alonso, J., Aguilera, C., Serrallonga, M., Pérez-Martín, J., Acebes, J. J., . . . Gili, J. (2003). Proton magnetic resonance spectroscopy (1 H MRS) of human brain tumours: assessment of differences between tumour types and its applicability in brain tumour categorization. European radiology, 13(3), 582-591.
• McBride, D. Q., Miller, B. L., Nikas, D. L., Buchthal, S., Chang, L., Chiang, F., & Booth, R. A. (1995). Analysis of brain tumors using 1H magnetic resonance spectroscopy. Surgical neurology, 44(2), 137-144.
• Nachimuthu, D. S., & Baladhandapani, A. (2014). Multidimensional texture characterization: on analysis for brain tumor tissues using MRS and MRI. Journal of digital imaging, 27(4), 496-506.
• Nagori, M., & Joshi, M. (2013). Methods and Algorithms for Extracting Values from MRS Graph for Brain Tumour Detection. IERI Procedia, 4, 331-336.
• Neugut, A. I., Sackstein, P., Hillyer, G. C., Jacobson, J. S., Bruce, J., Lassman, A. B., & Stieg, P. A. (2019). Magnetic Resonance Imaging‐Based Screening for Asymptomatic Brain Tumors: A Review. Oncologist, 24(3).
• Ramin, S. L., Tognola, W. A., & Spotti, A. R. (2003). Proton magnetic resonance spectroscopy: clinical applications in patients with brain lesions. Sao Paulo Medical Journal, 121(6), 254-259.
• Schuster, M., & Paliwal, K. K. (1997). Bidirectional Recurrent Neural Networks. IEEE transactions on Signal Processing, 45(11), 2673.
• Soares, D., & Law, M. (2009). Magnetic resonance spectroscopy of the brain: review of metabolites and clinical applications. Clinical radiology, 64(1), 12-21.
• Tate, A. R., Underwood, J., Acosta, D. M., Julià‐Sapé, M., Majós, C., Moreno‐Torres, À., . . . Murphy, M. M. (2006). Development of a decision support system for diagnosis and grading of brain tumours using in vivo magnetic resonance single voxel spectra. NMR in Biomedicine, 19(4), 411-434.
• Tsolaki, E., Svolos, P., Kousi, E., Kapsalaki, E., Fountas, K., Theodorou, K., & Tsougos, I. (2013). Automated differentiation of glioblastomas from intracranial metastases using 3T MR spectroscopic and perfusion data. International journal of computer assisted radiology and surgery, 8(5), 751-761.
• Vicente, J., Fuster-Garcia, E., Tortajada, S., García-Gómez, J. M., Davies, N., Natarajan, K., . . . Monleón, D. (2013). Accurate classification of childhood brain tumours by in vivo 1H MRS–a multi-centre study. European Journal of Cancer, 49(3), 658-667.
• Vieira, B. H., Santos, A. C. d., & Salmon, C. E. G. (2017). Pattern recognition of abscesses and brain tumors through MR spectroscopy: Comparison of experimental conditions and radiological findings. Research on Biomedical Engineering, 33(3), 185-194.
• Weis, J., Ring, P., Olofsson, T., Ortiz‐Nieto, F., & Wikström, J. (2010). Short echo time MR spectroscopy of brain tumors: grading of cerebral gliomas by correlation analysis of normalized spectral amplitudes. Journal of Magnetic Resonance Imaging, 31(1), 39-45.
• WHO. (2018). Cancer. Retrieved from https://www.who.int/news-room/fact-sheets/detail/cancer
• Wu, Y., Schuster, M., Chen, Z., Le, Q. V., Norouzi, M., Macherey, W., . . . Macherey, K. (2016). Google's neural machine translation system: Bridging the gap between human and machine translation. arXiv preprint arXiv:1609.08144.
• Zhang, X.-Y., Xie, G.-S., Liu, C.-L., & Bengio, Y. (2016). End-to-end online writer identification with recurrent neural network. IEEE Transactions on Human-Machine Systems, 47(2), 285-292.