2.2. Control Unit

The control unit (CU) is responsible for managing all the subsystems that comprise the demonstrator. This CU is a computer based on an Intel^® Core™ i7-4770k 3.5 GHz quad-core processor, with 8 GB of Random Access Memory (RAM) and a high-capacity 512 GB solid-state drive with write speeds exceeding 500 MB/s. Specific software was developed to manage and integrate the different elements that conform the acquisition platform, allowing the user to perform the HS image acquisition in an easy and effective way. Furthermore, the CU is in charge of executing the HS brain cancer detection algorithm together with the hardware accelerator in order to finally present the tumor boundary prediction.

Customized software for image acquisition was developed due to the need to automate and accelerate the capture of both HS cameras of the system. The simplification of the acquisition procedure ensures easy interaction of the user with the system as well as reduced time needed to capture the HS images during neurosurgical procedures.

To develop this software, three different software development kits (SDKs) were integrated, belonging to the two HS cameras and the stepper motor controller. Figure 4a shows the HS image acquisition software flow diagram for the capturing procedure. Firstly, after running the program, the scanning platform is initialized, detecting and establishing the absolute zero of the motor position. Then, the platform is positioned at the center of the scanning area. Taking into account the x-size value of the capturing area established by the user through the graphical user interface (GUI), the scanning platform is moved to the initial position. The VNIR capturing process is performed starting from the right to the left of the platform with the stepper motor speed fixed to 3 mm/s. This speed is calculated according to the pixel size (0.1287 mm and 0.48 mm for the VNIR and NIR cameras, respectively) and the frame rate of the camera (90 fps and 100 fps for the VNIR and NIR cameras, respectively). When the VNIR capture is done, the stepper motor stops at the final position, waits a few milliseconds to stabilize the system structure, and fixes the speed to 5 mm/s. Then, the NIR capturing process begins. This capture is performed starting from the left to the right of the platform. After that, the stepper motor moves the scanning platform to the central position. Then, the synthetic RGB images of both HS cubes are generated by selecting three bands that correspond with red (708.97 nm), green (539.44 nm), and blue (479.06 nm) colors for the VNIR image, and three bands of the NIR cube to generate a false color RGB image (red: 1094.89 nm, green: 1247.44 nm and blue: 1595.45 nm). These bands are selected to maintain the compatibility with the original software (Hyperspec^® III software, Headwall Photonics Inc., Fitchburg, MA, USA) provided by the camera manufacturer. Using this technique for the acquisition process, a speedup of 3× with respect to the original software is achieved. The maximum image size provided by the system is 1004 × 1787 pixels (129 × 230 mm) for the VNIR image, and 320 × 479 pixels (153 × 230 mm) for the NIR image, with spatial resolutions of 128.7 µm and 480 µm, respectively.

(a) HS image acquisition software flow diagram; (b) HS image acquisition user interface (and the RGB representations of each HS cube) being used during a neurosurgical intervention at the University Hospital Doctor Negrin of Las Palmas de Gran Canaria (Spain).

Figure 4b shows the acquisition system being used during a neurosurgical operation and the RGB synthetic images of the captured HS cubes (VNIR and NIR) where their image sizes and relative spatial resolutions can be seen. The time employed by the system to obtain the maximum size image using the manufacturer’s software is ~240 s for the VNIR image and ~140 s for the NIR image. However, employing the acquisition software developed in this work, the acquisition time for the maximum image size is reduced to ~80 s and ~40 s for the VNIR and NIR cameras, respectively.