Artificial intelligence in diagnostic neurosurgery—a perspective using artificial intelligence tools

As highlighted in the artificial intelligence (AI) Index Report 2024 (1), AI has demonstrated its ability to enhance productivity and elevate work quality. In neurosurgery, AI has potential for outcome prediction, providing insights to help decision making and interpretation of radiological, pathological and electrophysiological data.

This study focuses on AI in diagnostic neurosurgery, employing a methodology distinct from traditional systematic reviews and using an AI tool to identify emerging trends in the field. It also provides insights into potential future developments based on the study’s findings. While “diagnostic neurosurgery” itself does not directly encompass neuropathology, the two fields are closely related, and we have included neuropathology in our study.

A literature search was conducted through the National Library of Medicine website using the PubMed database. The search query “(Neurosurgery) AND (artificial intelligence) AND ((Diagnosis) OR (Diagnostic))” without specifying language was employed to identify relevant publications indexed before December 31, 2024.

A Python script (version 3.12.2) was utilized to extract the year of publication, title, and abstract (if available) from each retrieved record. The title and abstract of each record were joined together and were analysed for noun phrases, using TextBlob module for Python (2). Subsequently, each record was allocated to three diagnostic categories: radiology, pathology, or electrophysiology. This categorization was based on selected terms/phrases in Table 1. Emerging trends were derived from the extracted nouns based on sentiment analysis, rather than the frequency of the noun phrases (3).

One thousand two hundred seventy out of 1,304 initially retrieved records were included in the study due to the missing values from the remaining records. Figure 1 shows the increasing annual frequency of the total publications and in each diagnostic categories of radiology, pathology and electrophysiology from 2018 to 2024. Radiology leads the way, with the pathology category quickly gaining ground.

Figure 1 Annual frequency of the total publications and in each diagnostic categories of radiology, pathology, and electrophysiology from 2018 to 2024. AI, artificial intelligence.

Based on the extracted noun phrases, emerging trends in three other categories related to aneurysms (seach term ‘aneurysm’), spinal pathologies (search term ’spinal’) and glial tumours (search term ‘glio-’) were recognized. Figure 2 shows the annual frequency of publications in each of the categories of ‘aneurysms’, ‘glial’ and ’spinal’. While “glial’ leads the way, ’spinal’ category has gained ground significantly in 2023–2024.

Figure 2 Annual frequency of publications in each of the categories of aneurysms, glial, and spinal.

For aneurysms, the primary emerging trend appears to be the development of computed tomography angiography-based AI models for detecting cerebral aneurysms, aimed at serving as a radiologist assistant and/or a screening tool. One of the major technical hurdles in AI-powered, image-based diagnostic neurosurgery is accurately segmenting regions of interest, whether normal or pathological, within medical imaging. AI-based automatic segmentation (4) and AI-assisted pathological diagnosis appear to be the emerging trends for the future.

There is an increasing practical demand for AI models in intraoperative neuropathological diagnosis, with trends indicating a future focus on combining stimulated Raman histology, surgical microscope videos/photos, and preoperative imaging, as well as using AI algorithms to enhance precision in molecular pathology and analyzing histopathological images.

In spinal category, the focus of AI has been on spinal measurements in surgical planning for a variety of spine procedures. Performing these measurements on full-length imaging eliminates distortions that can occur with stitched images. However, these images take radiologists significantly longer to read than conventional radiographs. There are challenges in using AI for diagnosing fractures above T9, differentiating between acute and subacute injuries, and identifying underlying pathologies.

AI requires large datasets for training. Although there are publications exploring the use of AI in neurosurgical clinical diagnosis and decision-making, it is not surprising that the most significant growth has occurred in fields like radiology, which involve large imaging datasets. There is a significant role for AI in analysis and correcting errors in the large data collected by wearable devices and neurophysiological monitoring devices such as scalp electroencephalography and magnetoencephalography. Using AI pattern recognition models, specific patterns can be identified through portable AI-powered devices and passed directly to a chatbot, eliminating the need for apps or a touchscreen.

The expanding role of AI in diagnostic neurosurgery underscores the need to standardize large data formats and promote collaboration between medical centers at both national and international levels. However, there are ethical challenges that must be addressed, including data privacy concerns and the lack of diagnoses in underserved and underrepresented populations, which heightens the risk of systemic bias. Additionally, the involvement of third parties, outside the physician-patient relationship, in the use of AI, as well as the inability to respond to unanticipated events during diagnostic procedures, presents further challenges for AI in diagnostic neurosurgery.

Our study had limitations. It treats all related publications as equally valuable. The keywords used might not encompass all aspects of diagnostic neurosurgery and there are overlaps between different categories that may impact the accuracy of the statistics provided. While more advanced phrase extraction methods than TextBlob are available (5), we chose this approach for its simplicity, considering the limited amount of training text.

Acknowledgments

None.

Footnote

Provenance and Peer Review: This article was commissioned by the editorial office, AME Surgical Journal. The article has undergone external peer review.

Peer Review File: Available at https://asj.amegroups.com/article/view/10.21037/asj-25-32/prf

Funding: None.

Conflicts of Interest: Both authors have completed the ICMJE uniform disclosure form (available at https://asj.amegroups.com/article/view/10.21037/asj-25-32/coif). B.E. serves as an unpaid editorial board member of AME Surgical Journal from August 2024 to December 2026. The other author has no conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. Nestor M, Fattorini L, Brynjolfsson E, et al. The AI Index 2024 Annual Report. Institute for Human-Centered AI, Stanford University; 2024.
2. Loria S. Textblob Documentation. Release 015 2018;2:269.
3. Peña-Cáceres O, Silva-Marchan H, Espinoza-Nima R, et al. Comparison of the VADER and TextBlob Models in Sentiment Analysis. In: Ibáñez DB, Gallardo-Echenique E, Siringoringo H, et al. editors. Communication and Applied Technologies. ICOMTA 2024. Singapore: Springer Nature Singapore; 2024.
4. Cekic E, Pinar E, Pinar M, et al. Deep Learning-Assisted Segmentation and Classification of Brain Tumor Types on Magnetic Resonance and Surgical Microscope Images. World Neurosurg 2024;182:e196-204. [Crossref] [PubMed]
5. Eftekhar S, Eftekhar B. Neurosurgical literature classification - Evaluation of three automated methods and time trend analysis of the literature. Heliyon 2024;10:e26831. [Crossref] [PubMed]