Development of Bleeding Artificial Intelligence Detector (BLAIR) System for Robotic Radical Prostatectomy

Abstract

Background: Addressing intraoperative bleeding remains a significant challenge in the field of robotic surgery. This research endeavors to pioneer a groundbreaking solution utilizing convolutional neural networks (CNNs). The objective is to establish a system capable of forecasting instances of intraoperative bleeding during robot-assisted radical prostatectomy (RARP) and promptly notify the surgeon about bleeding risks.

Methods: To achieve this, a multi-task learning (MTL) CNN was introduced, leveraging a modified version of the U-Net architecture. The aim was to categorize video input as either "absence of blood accumulation" (0) or "presence of blood accumulation" (1). To facilitate seamless interaction with the neural networks, the Bleeding Artificial Intelligence-based Detector (BLAIR) software was created using the Python Keras API and built upon the PyQT framework. A subsequent clinical assessment of BLAIR's efficacy was performed, comparing its bleeding identification performance against that of a urologist. Various perioperative variables were also gathered. For optimal MTL-CNN training parameterization, a multi-task loss function was adopted to enhance the accuracy of event detection by taking advantage of surgical tools' semantic segmentation. Additionally, the Multiple Correspondence Analysis (MCA) approach was employed to assess software performance.

Results: The MTL-CNN demonstrated a remarkable event recognition accuracy of 90.63%. When evaluating BLAIR's predictive ability and its capacity to pre-warn surgeons of potential bleeding incidents, the density plot highlighted a striking similarity between BLAIR and human assessments. In fact, BLAIR exhibited a faster response. Notably, the MCA analysis revealed no discernible distinction between the software and human performance in accurately identifying instances of bleeding.

Conclusion: The BLAIR software proved its competence by achieving over 90% accuracy in predicting bleeding events during RARP. This accomplishment underscores the potential of AI to assist surgeons during interventions. This study exemplifies the positive impact AI applications can have on surgical procedures.

Keywords: artificial intelligence; complications; prostate cancer; robotics.

Figure 1

Architecture of the employed CNN. Multi-channel feature maps are represented by each box. The box top displays the number of channels, while its lower left border displays the x-y dimensions. The backbone is represented by blue boxes, the first branch for semantic segmentation is shown by green boxes, copied feature maps are indicated by gray boxes, and the second branch for event detection is indicated by purple boxes. The various operations are shown by the arrows.

Figure 2

The BLAIR application interface. (a) No bleeding prediction by the Core NN. When repeated prediction with a small confidence percentage occurs in a time frame, the alert level in the left bar increases (b) up to the point when the bleeding event occurs, as it can be seen in the dotted circle (c). The gravity of the event does not change substantially the alert level (d).

Figure 3

Training and validation metrics are trends. Training loss (a) and validation accuracy for the event detection branch (b).

Figure 4

Examples of BLAIR software performance evaluations. (a) The software correctly predicts the bleeding (true positive); (b) the software wrongly predicts a bleeding that did not occur (false positive); (c) the software fails to predict a bleeding (false negative).

Figure 5

Clinical evaluation of BLAIR software performance. (a) Density plot of the delta time (BLAIR vs. human) in terms of seconds. The x-axis represents the difference in time between the BLAIR and clinical (human) measurements, and the y-axis represents the density of data points in that range. It reports a histogram of the collected data, together with a KDE (Kernel Density Estimation) plot, i.e., a smoothed representation of the data’s probability density function. (b) Plot of the MCA dimension. (c) Plot of the MCA dimensions in terms of the individuals under analysis.