Instrument Identification to improve OR Scheduling

Abstract

Hospitals are under an increased financial pressure, due to a significant growth in demand for healthcare, that is caused by an aging population. The operating rooms account for the largest costs of a hospital, and thus require a cost-efficient management. However, scheduling the operating rooms is a complex task that involves the estimation of surgery duration, which is often inaccurate. Delays in surgery duration, that are not accounted for in the estimation during scheduling, often result in upcoming surgeries being cancelled. Hence, the operating theatres are most of the time not fully utilized, which comes with great costs for the hospital. Poor communication of information about the progress of ongoing surgeries to the operating room (OR) schedulers prevents them from effectively (re)-scheduling accordingly. The OR personnel can not be assigned to additional tasks, and thus, there is a need for automated monitoring of the progress of ongoing surgeries. The instrument usage during surgeries is shown to provide relevant information about the progress of surgeries, using a Random Forest model to estimate surgical phases from intra-operative data.In this thesis, a Radiofrequency Identification (RFID) setup is used to detect instruments, equipped with RFID-tags, that are used by the surgeon during a Total Extraperitoneal Laparoscopic Hernia Repair (TEP) procedure at Reinier de Graaf Gasthuis (RdGG). The setup is evaluated on applicability into a surgical setting, which involves regulatory aspects, acceptance by the OR-personnel, and performance. A pilot study, of which the results are incorporated into realizing actual in-vivo tests, is performed in a near-surgical setting to eval- uate the extent to which the RFID-setup is applicable in a real surgical setting. During the pilot study, a total of 10 procedures is mimicked of which the data, that is extracted from the RFID-setup, is used to generate a Random Forest model that can estimate the surgical phases. The Random Forest model is used to evaluate the performance of the RFID-setup. After complying with all regulations to perform in-vivo testing, a total of three in-vivo tests is executed to evaluate the applicability of the RFID-setup in a real surgical setting. Lastly, a first step has been taken to extend the research to a greater range of surgery types.The RFID-setup has shown the ability to identify instruments, that are equipped with RFID-tags, during surgery. All regulations to perform in-vivo tests are complied with, and no discomfort of the RFID-setup is experienced by the OR-personnel during the in-vivo tests. Hence, future in-vivo testing is warranted to optimize the setup. The Random Forest model can estimate the phases of surgeries (38% overall accuracy), based on instrument usage data that is acquired through an RFID-setup. The current setup, however, needs to be optimized to improve the detection ability. So far, the first step to make the method of instrument identification applicable to a broader range of surgeries has resulted in a prototyped attachment design to equip instruments with RFID-tags. The attachment is designed in such a way that both the attachment and the RFID-tags are interchangeable between instruments. Further development is needed for the design to comply with a selection of regulations to perform in-vivo testing.