The PROST ERC-PoC project aims at applying the technologies developed by the ERC-AdG project ARS to demonstrate the scientific, medical, and economic potential of using methods of artificial intelligence within a robotic embodiment to reduce human errors in medical procedures.
The chosen procedure was prostate biopsy, which is the golden standard to verify the presence of prostate cancer (PC) in a patient. The procedure is routinely carried out in hospital worldwide and it is still affected by more than 30% diagnostic error, due primarily to human cognitive and dexterity errors. The former are generated during the visual analysis of the prostate Magnetic Resonance Images (MRI), by overlooking potential pathological areas, and the latter occur during the execution phase, by the inaccurate insertion of the biopsy needle in the patient prostate under Ultra Sound (US) guidance.
PROST addressed these two error sources, by performing the detection of biopsy targets with the assistance of machine learning algorithms during MRI analysis, and by developing trajectory planning and robotic execution to precisely position the needle towards the target. Then, we apply fusion techniques to map the lesions identified in the MRI images to the real time prostate images acquired using an US echographer. US images provide also visual feedback to the physician during needle insertion by supporting the 3D reconstruction of the prostate and of the actual needle trajectory. By addressing the above error sources, we have reached Autonomy Level 1, i.e. this device is providing cognitive and manual assistance to the physician to improve the outcome of the prostate biopsy. The physician has the final decision on accepting the PC targets identified by PROST and she/he manually inserts the needle and triggers the acquisition of the tissue samples.
To verify the potential impact of this technology, we have carried out laboratory tests, financial analysis and regulatory examinations. After completing the PROST prototype, we first carried out extensive laboratory tests to assess the procedure accuracy. The results were promising, and consistently matched the accuracy needed to sample lesions smaller than 5 mm of diameter, which is the current resolution of MRI PC analysis. This phase has been documented in a paper published on the International Journal of Computer Assisted Radiology and Surgery (IJCARS). Subsequently, we received two approvals from the Ethical Committee of the Verona Hospital: the first to carry out usability and accuracy tests on cadavers, and the second to acquire patient’s MRI and US images to train the machine learning algorithms for target identification. The results of the cadaver tests will be soon published.
The feasibility of developing a product based on PROST was analysed by developing a Business Plan (BP) of the potential start-up. We have also identified the documents that describe the main steps and requirements of the certification process to acquire the CE mark for PROST as a medical device. The BP was presented to potential investors and the preparation of the certification documents is under way.
It must also to be noted that the target price of the potential commercial product aims at allowing all urologists to acquire the device for use in their ambulatory practice because the final device will be adaptable to the US echographers already available at the physicians’ offices. This will further redue the financial impact of this approach on the physicians and on the National Health Care systems as well, while increasing the overall diagnostic accuracy and patient quality of care.
