Viewpoint

Using Neural Networks for Tumor Classification during Surgery

by Alyona Ivanova

Graphic design by Brendan Lazar

Central nervous system (CNS) tumors, especially in children, are considered highly lethal cancers.1 The primary approach to treatment involves surgically removing the tumor; a procedure that necessitates a delicate balance between maximizing resection extent and minimizing the risk of neurological damage and associated complications.2 Unfortunately, surgeons often lack precise information about the tumor type before surgery. The current standard practice relies on pre-operative imaging and intraoperative histological analysis, but their conclusiveness is not guaranteed and can sometimes be inaccurate.3

In response to this diagnostic uncertainty, there is a growing exploration of innovative technologies to augment surgical decision-making processes. Notably, the integration of neural network (NN) technologies for tumor classification during surgery emerges as a promising avenue. These advanced computational models have the potential to revolutionize the field by providing real-time insights into the molecular characteristics of tumors. By leveraging rapid sequencing for real-time analysis of genomic DNA and employing sophisticated NNs there is an opportunity to enhance the accuracy and efficiency of tumor classification, facilitating more informed and precise surgical interventions.5-7 For instance, the developed Sturgeon model is a patient-agnostic transfer-learned neural network that enables molecular subclassification of CNS malignancies based on sparsity profiles.4  This introduction sets the stage for an exploration into the benefits and challenges associated with utilizing NN in the intricate landscape of CNS tumor surgery.

The use of  NNs is just one application of artificial intelligence (AI) technologies in the operating room (OR). AI use during surgery for tumor classification offers several potential benefits, with the greatest being real-time diagnosis. AI algorithms can rapidly analyze data, providing real-time classification of tumors. This enables surgeons to make informed decisions on the spot, potentially improving the efficiency of the procedure.3 Furthermore, machine learning models can analyze complex datasets, including genomic and molecular information, with a high degree of accuracy. This can lead to more precise tumor classification compared to traditional methods, reduce the likelihood of misdiagnosis, and improve treatment planning.4 AI algorithms can also learn from a large volume of data, continually improving their accuracy and performance over time. This adaptive nature can contribute to ongoing advancements in tumor classification and surgical decision-making.4 

Precision medicine, also known as personalized medicine, involves tailoring medical treatment and interventions to the individual characteristics of each patient.9 In the case of tumor classification, precision medicine aims to customize treatment plans based on the unique molecular and genetic profile of the tumor. This personalized approach may enhance the effectiveness of therapies. Moreover, accurate intraoperative tumor classification can help surgeons achieve optimal resection during the initial surgery, potentially reducing the need for additional surgeries. This can lead to quicker recovery times and improved patient outcomes.4 Finally, by providing timely and accurate information about the tumor type, AI can assist surgeons in striking a balance between maximizing the extent of resection and minimizing the risk of neurological damage.4 This is particularly crucial in CNS surgeries.

While these benefits are promising, it’s important to note that the integration of AI technologies into surgical workflows requires careful validation, ethical considerations, and collaboration between medical professionals and AI experts to ensure patient safety and the effectiveness of these technologies in clinical settings.10 AI models heavily rely on the quality and representativeness of the training data. If the training data is biased or lacks diversity, the AI algorithm may produce biased results or struggle to generalize to diverse patient populations.  While an AI model may perform well on a specific dataset, its performance on new, unseen data is crucial.10 Establishing the reliability of these models across diverse patient populations and clinical settings requires rigorous validation. Additionally, many AI models, especially complex deep learning models, are often considered “black boxes”11 because their decision-making processes are not easily interpretable. This lack of transparency can be a concern for medical professionals seeking to understand and trust the AI-generated recommendations. However, striking the right balance between the expertise of healthcare professionals and the capabilities of AI systems is crucial. Over-reliance on AI or the neglect of clinician expertise can lead to suboptimal patient care.

The use of AI in healthcare also raises ethical issues related to patient privacy, consent, and the responsible use of sensitive medical data.12 Legal frameworks and guidelines must be in place to address these concerns and protect patient rights. Regulatory frameworks for the approval and deployment of AI technologies in healthcare are still evolving.12 Navigating regulatory hurdles and ensuring compliance with standards and guidelines can be a complex process.

Finally, incorporating AI technologies into existing surgical workflows may pose challenges. Surgeons and healthcare professionals need to adapt to these new tools seamlessly, and integration must not disrupt the efficiency and effectiveness of clinical practices.10 Furthermore, implementing AI technologies in healthcare settings may require significant financial investments in infrastructure, training, and maintenance.10 This could be a barrier for some healthcare institutions, particularly those with limited resources.

In conclusion, the integration of AI, particularly NN technologies, in the realm of CNS tumor surgery holds immense promise for revolutionizing diagnostic precision and treatment personalization. The challenges and benefits associated with employing AI in this critical medical domain underscore the transformative potential of technology in enhancing patient outcomes. In essence, the journey toward incorporating AI technologies into CNS tumor surgery represents a dynamic intersection of innovation, medical expertise, and ethical responsibility. Ongoing research, collaborative efforts, and the establishment of robust frameworks will be instrumental in navigating the complexities, fostering trust, and ultimately realizing the transformative potential of AI in improving patient care and outcomes in the challenging landscape of neurosurgical interventions.

  1. Cohen AR. Brain tumors in children. N. Engl. J. Med. 2022;386(20):1922-1931.
  2. Yong RL, Lonser RR. Surgery for glioblastoma multiforme: striking a balance. World Neurosurg. 2011;76(5):528-530.
  3. Djirackor L, Halldorsson S, Niehusmann P, et al. Intraoperative DNA methylation classification of brain tumors impacts neurosurgical strategy. Neurooncol Adv. 2021;3:vdab149.
  4. Vermeulen C, Pagès-Gallego M, Kester L, et al. Ultra-fast deep-learned CNS tumour classification during surgery. Nature. 2023;622:842-849. https://doi.org/10.1038/s41586-023-06615-2
  5. Gorzynski JE, Goenka SD, Shafin K, et al. Ultrarapid nanopore genome sequencing in a critical care setting. N Engl J Med. 2022;386:700-702.
  6. Sagniez M, Simpson SM, Caron M, et al. Real-time molecular classification of leukemias. Preprint at medRxiv. 2022 https://doi.org/10.1101/2022.06.22.22276550  
  7. Guo C, Pleiss G, Sun Y, et al. On calibration of modern neural networks. In Proc. 34th International Conference on Machine Learning, Vol. 70 (eds Precup D, Teh YW). Proceedings of Machine Learning Research. 2017;1321-1330.
  8. Kuschel LP, Hench J, Frank S, et al. Robust methylation-based classification of brain tumors using nanopore sequencing. Preprint at bioRxiv. 2021 https://doi.org/10.1101/2021.03.06.21252627
  9. Naithani N, Sinha S, Misra P, et al. Precision medicine: Concept and tools. Med J Armed Forces India. 2021;77(3):249-257. https://doi.org/10.1016/j.mjafi.2021.06.021  
  10. Kelly CJ, Karthikesalingam A, Suleyman M, et al. Key challenges for delivering clinical impact with artificial intelligence. BMC Med. 2019;17(1):195. Published 2019 Oct 29. https://doi.org/10.1186/s12916-019-1426-2
  11. Rudin C, Radin J. Why Are We Using Black Box Models in AI When We Don’t Need To? A Lesson From an Explainable AI Competition. Harvard Data Science Review. 2019;1(2). https://doi.org/10.1162/99608f92.5a8a3a3d.
  12. Jobin A, Ienca M, Vayena E. The global landscape of AI ethics guidelines. Nat Mach Intell. 2019;1:389-399. https://doi.org/10.1038/s42256-019-0088-2.