A Wearable Infrared Thermographic Imaging System for Real-Time Intraoperative Brain Tumor Visualization: A Proof-of-Concept Study

Intraoperative tumor differentiation remains a challenge in neurosurgical procedures. Distinguishing between malignant and normal tissue during surgery using very cost-effective smart glasses and obtaining exact guidance is very challenging [1, 2]. Many studies employ infrared (IR) cameras to measure the temperatures of different parts of the human body, especially in breast, dermatological, and heart diseases [3, 4]. This study presents the design and development of a smart IR glasses platform for real-time characterization of malignant and normal brain tissues during surgery. The system, equipped with an AMG IR sensor, captures high-resolution thermal maps displayed directly on the glasses, providing the surgical team with immediate, non-invasive thermal feedback. Highly malignant with a slow growth rate, gliomas are tumors that arise from glial cells in the brain or spinal cord and have four grades from low to high malignancy. In children, a pilocytic astrocytoma is the most common glioma [5, 6]. It can be characterized by frequent B-Raf proto-oncogene, serine/threonine kinase gene (BRAF) alterations, circumscribed growth, and a favorable prognosis. It is divided into clinically indolent low-grade tumors and high-grade tumors with aggressive behavior [7, 8]. The following case study represents the latter category. Multimodal imaging is commonly used to diagnose brain tumors and provide 3D images of the brain, such as computed tomography (CT), magnetic resonance imaging (MRI), and positron emission tomography (PET) [9-11]. All these modalities are used preoperatively and do not provide real-time images in the operating room [12-14]. IR sensors can detect heat emitted from brain tumors, especially malignant ones. The increased blood flow of the cancerous cells and a high metabolic rate lead to elevated cell temperature to a higher level than surrounding tissue [15, 16]. The average temperature of healthy brain tissue reported ranged between 28 and 33 ℃ [17-19]. Koga et al. [20] in Japan were the first to use an IR camera intraoperatively in brain surgery. They proved that intracranial neoplasms exhibited thermal contrast between different brain regions. They tested the thermal microenvironments of six brain tumors by using a low-resolution IR camera. Their results proved that brain tumor temperatures differed in relation to their types. Gorbach et al. [21] used 34 patients with primary and metastatic brain tumors to conduct one of the earliest clinical series by using IR cameras intraoperatively. The study confirmed that temperature gradients correlated with pathological lesions. That offering early evidence that tumor‑related perfusion changes could be visualized via IR imaging [22]. Steiner et al. [23] explored IR spectroscopy for brain tumor recognition. They analyzed spectroscopic signatures from resected glioblastoma specimens and non‑tumor brain tissue. Their classification model correctly differentiated tumor from normal tissue with high accuracy. That emphasizing that IR spectral biomarkers may contribute to intraoperative tissue classification beyond pure thermal maps [24]. Menezes et al. [25] explored real-time IR thermography for brain tumor classification during surgery. Their study of 35 patients showed that IR imaging could effectively tell apart malignant and benign tissues. They also mentioned the potential of combining IR systems with Augmented Reality to show thermal data directly in the surgeon’s view, which could make surgeries more efficient [23, 25]. Da Costa et al. [26] proved that using IR thermography to localize intracranial lesions was feasible. They examined residual tumor detection during vascular and epilepsy surgeries. They demonstrated that IR cameras can localize lesions within up to 2 cm in depth with high sensitivity and accuracy. The residual tumor exhibited nonspecific static and characteristic dynamic thermographic maps. The reported temperature difference is about 0.5–2.0 ℃ between the brain tumor margin and the healthy cortex. In accordance with the literature introduced above, several limitations were faced by the researcher. That includes low spatial resolution of thermal sensors, sensitivity to environmental conditions, the influence of blood flow and tissue perfusion on temperature distribution, and challenges in accurately correlating thermal patterns with anatomical structures [27]. Therefore, designing and developing non-invasive, cost-effective glasses based on IR technology is very challenging. It provides real-time monitoring of brain resection in the operating room. This work was validated by performing surgery in the Department of Neurosurgery in Ghazi Al-Hariri Surgical Specialty Hospital, Medical City Campus, Baghdad, Iraq. A case study on 14 years old patient with a glioma tumor was performed. Evaluation of the smart glasses' performance was conducted.

3.1 Results

Figure 4 shows a pre-incision IR image from behind the scalp. It shows the tumor location very clearly with a hot spot and a core temperature of 40.1 ℃. The surrounding area has a decreased temperature. Alongside is a visible image of the same view. The temperature shown guides the surgeon to get into the tumor directly without needing another approach.

4.png

Figure 4. Pre-incision infrared (IR) and visible image

5.png

Figure 5. Infrared (IR) and visible image after removing the cortex

6.png

Figure 6. Infrared (IR) image of the biopsy after being removed from the brain

In Figure 5, the surgeon removed the cortex and exposed the brain to reach the tumor. Again, the core temperature here is higher than the surrounding tissue, which makes it a guide during the surgery. The tumor temperature was 36.2 ℃ with an emissivity of 0.98. This is used because most of the brain tissue is composed of water, so the emissivity of water is 0.89. The right image in Figure 5 shows the visible image.

Figure 6 shows the biopsy temperature taken immediately after being extracted from the brain. It helps the surgeon determine whether the removed tissue is malignant or normal. In this image, the temperature of the biopsy is approximately 39 ℃, which is very high in comparison with normal brain tissue. The temperature of normal brain tissue ranges from 28 to 33 ℃.

3.2 Experimental evaluation

Several methods are used to evaluate the performance of the smart glasses mentioned below:

1. A histopathological report after surgery was performed on the excised tumor tissue. This method was used to validate the thermal map patterns observed during surgery. The observed intraoperative thermal patterns were compared qualitatively with the lesion location. MRI and the surgical field identified these locations. demonstrating correspondence between the thermal observations and the confirmed pathological region. This is shown in Figure 7, alongside the MRI image of the patient preoperatively.

7.png

Figure 7. Magnetic resonance imaging (MRI) of the patient with a histopathology report

2. The proposed system maintains a compact and energy-efficient architecture. It is suitable for wearable intraoperative applications while maintaining sufficient operational time for clinical procedures. Table 4 shows the power consumption of each component used. The total estimated system power consumption is approximately 1.60 watts.

Table 4. Power consumption

3. To measure the temperature difference between the tumor and the surrounding cortex, use the mean temperature equation

$\Delta T=T_{\text {tumor}}-T_{\text {normal}}$        (1)

where,

•T_tumor = average temperature of the tumor region.

•T_normal = average temperature of adjacent cortex.

where, the average tumor temperature is 37.55 ℃, and the normal average was 32.9 ℃. The observed thermal contrast between tumor and surrounding cortex (4.65 ℃) is consistent with previously reported intraoperative thermography findings in malignant brain tumors. This finding is consistent with previously reported intraoperative IR image observations. The malignant brain tissue may show altered high temperature compared with adjacent normal tissue due to differences in tissue metabolism, vascularity, and physiological activity.

4. The contrast-to-noise ratio (CNR) between the suspected tumor region and surrounding cortex was calculated:

$C N R=\frac{\left|T_{\text {tumor}}-T_{\text {normal}}\right|}{\sigma_{\text {normal}}}$        (2)

where, σ_normal represents the standard deviation within the normal cortical region of temperature measurements. A higher CNR indicates improved separation between the thermal signal of the suspected tumor region and the surrounding tissue.

5. Qualitative assessment of anatomical correspondence was performed. It compared preoperative MRI findings with the intraoperative surgical field and IR images. The lesion location identified on MRI was reviewed by the surgeon with respect to the exposed surgical area. The corresponding thermal map observed during intraoperative imaging provided qualitative anatomical consistency among preoperative imaging, intraoperative observation, and thermal visualization.

6. Quantitative thermal assessment of intraoperative IR imaging was extracted from the IR image as mentioned in Table 5.

Table 5. Quantitative thermal assessment

7. The real-time performance of the proposed system was evaluated based on the measured acquisition and processing times of each component. The execution time of each processing stage was recorded, including thermal image acquisition, data transmission, image processing, and display updating. The total system latency was estimated to be approximately 120–140 ms. This corresponds to an effective frame rate of approximately 8 FPS (Frames Per Second). The overall latency was calculated as the cumulative delay of all processing stages. Component-wise analysis was performed to identify the contribution of each stage to the final system response. The achieved performance is mainly limited by the refresh rate of the AMG8833 thermal sensor. Nevertheless, this latency enables real-time thermal image visualization and monitoring. Power consumption was evaluated during normal operation by assessing the electrical requirements of the integrated components, including the IR sensor, processing unit, and display module. Figure 8 has been clarified to show the contribution of each component to the total latency.

8.png

Figure 8. Total system latency

The proposed system demonstrated the capability to perform real-time IR imaging of the subcortical tumor region. It performed at depths reaching approximately 10 mm, along with the adjacent cortical tissue. The AMG8833 sensor captured thermal variations, indicative of tumor metabolic activity and abnormal vascularization. These were clearly distinguishable from adjacent healthy tissue. The 1.69-inch LCD provided immediate visual feedback to the surgeon without the need to look away from the operative field. It highlights the practicality of integrating FPV glasses with dual imaging modalities for intraoperative use. The tactile switches enabled the surgeon to control display modes and calibration conveniently without interrupting the surgical workflow. Also, it illustrates the advantage of a hands-free, surgeon-oriented interface. The findings of this study suggest that a compact embedded platform, such as the RP2040 microcontroller, is capable of supporting real-time thermal data acquisition, processing, and visualization. Despite its limited computational power compared with larger systems, the platform demonstrated sufficient performance for operation in a controlled surgical environment. The combined visible and IR imaging provided a complementary perspective; visible imaging. This provided anatomical context, while thermal imaging highlighted areas of potential malignancy. This fusion may assist in identifying regions of thermal contrast. It could support intraoperative assessment of tumor boundaries. However, further validation using larger patient cohorts and quantitative pixel-wise thermal analysis is required. Although this was a single case study, the results indicate the feasibility of low-cost, portable wearable IR imaging systems in pediatric neurosurgery. Further studies with larger patient cohorts and higher resolution sensors are necessary to quantify the sensitivity and specificity of IR imaging. Additionally, integration with augmented reality overlays and automated thermal analysis could further enhance surgical precision and safety.