SPECTRA: Intraoperative Hyperspectral Imaging for Prostate Cancer

Prostate cancer (Pca) is the second most frequently diagnosed cancer in men worldwide with over 1.5 million new cases annually and a median age at diagnosis of 66 years. Despite the natural history being slow and clinically insignificant in most cases, its heterogeneity and high prevalence mean it remains the second leading cause of cancer-related death worldwide. Treatment options depend on cancer stage, risk stratification and patient preferences.

Radical prostatectomy (RP), which consists of the surgical removal of the prostate and the seminal vesicles, is one of the gold-standard curative treatment for localized diseases. RP techniques include open, laparoscopic, and robotic-assisted surgery. Despite recent advances in the robotic field, positive surgical margins (PSM) can occur up to 42 %. According to the International Society of Urological Pathology (ISUP), PSM is defined by tumor cells reaching the inked surgical margin of the prostatectomy specimen. PSM significantly reduces progression-free survival and is a known prognostic parameter for postoperative biochemical recurrence (BCR).

Current margin assessment relies on Intraoperative Frozen Section (IFS), the current gold-standard technique is the Neurovascular Structure-Adjacent Frozen-Section Examination (NeuroSAFE). It consists of an IFS analysis of tissue adjacent to neurovascular bundles. While this neurovascular structure sparing method improves postoperative erectile function and reduces PSM, its implementation is limited due to additional operating time (45-60 minutes), increased resource demands on pathology departments, and the inability to provide a whole specimen analysis.

Hyperspectral Imaging (HSI) combines conventional imaging (camera) with a spectrometer, providing both spatial and spectral information of the analyzed structures, enabling tissue differentiation based on structural properties. HSI uses reflectance spectroscopic imaging measurements, which involve a white light irradiation of the tissue and subsequently a recording of the remitted spectral intensities. The interaction between the incident light and the tissue modifies the light's spectral distribution with the reflected light containing information about the tissue composition. This technology has already been successfully applied in visceral surgery to quantify liver viability during ischemic and reperfusion phases. More recently, in Urology, HSI was successfully used as a predictive tool for early postoperative kidney graft.

Another clinical application of HSI is the ability to discriminate between normal and cancerous tissue as observed in gastric, head and neck (H&N), colorectal, brain and breast cancers. Similarly, in 2024, spectral data from Diffuse Reflectance Spectroscopy (DRS) showed promising results achieving an average sensitivity of 89% with a specificity of 82%, in 59 prostatectomy specimens, confirming the potential of spectroscopy. However, DRS approach comes with limitations, more particularly in spatial orientation because it requires multiple point-based measurements across the prostate surface with the inability to evaluate the entire tissue volume. In contrast, HSI provides spatially resolved spectral data across the entire field of view, enabling comprehensive tissue mapping. In 2012, the feasibility of HSI for PCa detection was evaluated in in-vivo mice and in pathological tissue slides, achieving 93% sensitivity and 97% specificity using a support vector machine algorithm and providing an estimated AUC-ROC of 0.9485. In this context, HSI provides a unique opportunity to assess fully a specimen. However, its application for real-time margin assessment during surgery remains unexplored. Recently, a 3D pipeline that co-registers multi-angle ex-vivo HSI with volumetric histopathology was implemented, enabling voxel-wise tumor segmentation and AR visualization in H&N specimens using convolutional neural networks (CNNs).