Cost on the tedious counting task
Take a guess, how many major surgeries take place around the world each year? According to statistics, the number is a staggering 310 million (Dobson, 2020)!
In order to ensure the safety for the patients who receive surgeries, surgical instrument has to be count before and after the surgery. Take another guess, how many man hours by professional medical staff have to spend on this check-in and check-out?
Estimate by local data, 2 minutes for counting during check-in, 3 minutes for matching the count sheet, and 4 minutes for counting again during check-out by 2 persons, each surgery requires 13 minutes for such work. It roughly equals 67 million labor hours around the world!
This is a huge number. Think about it, if that 67 million hours by professional medical staff can be redeployed to other life-saving jobs instead of just the counting task, how many extra lives can be saved each year?
Remember, this number may be higher because for some major surgeries more rounds of counting of surgical instruments may be needed, meaning more nurses and supporting staff members have to count them in and out of the operation theatre for sterilization.
Blueinno's solution: An AI automatic system for counting surgical instruments
That's why Blueinno AI invented an AI solution for automating the counting of surgical instruments, as part of their endeavor of developing the "smart hospital", a vision for advancing the medical industry into the future.
"Putting so many manpower onto the tedious task of counting surgical instrument is just a waste of the skills and talents of the medical staff", said Monica Leung, founder and director of Blueinno AI, "this is a kind of misplacement of human capitals in society. We developed such system in purpose for optimizing our medical system".
Blueinno's solution: counting surgical instruments automatically
Blueinno employed artificial intelligence (AI), integrated with computer vision, to replace the manual counting tasks by breaking down which into three sub-tasks:
1. Recognition: what are those surgical instruments?
2. Localization: where is those surgical instruments?
3. Counting: How many of different types of those surgical instruments are there?
The technology behind the system.
The technology employed for the system "We started from collecting the users' expectations, followed by developing the overall architecture in terms of hardware (the front-end camera and a GPU) and software (the back-end AI model) for machine learning, which is a iterative process by feeding various types of surgical instruments as many as possible from different angles into our system, so as to optimize the system performance," shared by Dr Calvin Kam, the technical director of Blueinno AI.
In summary, the whole development process involves 6 stages:
- 1. Data collection
- 2. Data augmentation
- 3. Data annotation
- 4. Model training
- 5. Evaluation
- 6. Deployment
Better accuracy, shorter time
As a result, the system is able to almost completely replace the current counting task carried out by medical staff, with performance of nearly 98% of accuracy. Moreover, such counting takes at least 10 minutes to achieve in the past, now can be done in a matter of seconds.
"The most important thing is, our system can achieve stable performance all the time because machines will not get tired, and their performance will only get better with more data accumulated", further shared by Monica.
The system demonstration
Save more lives, less cost
Such high accuracy and stable performance can also prevent medical accidents of leaving surgical instrument in patients' bodies from happening. In fact, this is a global problem for surgeons an hospital around the world. For instance, according to estimate by National Center for Biotechnology Information (NCBI), there are 0.3 to 1.0 piece of surgical instruments left inside patients per 1,000 abdominal operations, or 1,500 of such incidents for the United States per year.
Furthermore, such retention incidents also spin off significant financial impacts. According to Nate Miersma, director of Stryker, a medical device and equipment manufacturer, A single incident can add up to the extra cost of US $600,000, including the legal cost and corrective measures for rectification.
Therefore, automatic counting of surgical instruments can bring various benefits to the society, including better medical quality at less expense, reducing total cost to society, better deployment of our medical resources and talents, shortening time for surgery by excluding time for manual counting. In short, save more lives at less cost.
The system deployment in the real environment
Winning AI Challenge and system deployment in the Hong Kong hospital
Such an innovative AI and computer vision system let Blueinno won the overall winner in AI Challenge, the competition organised by Hospital Authority of Hong Kong and Hong Kong Science & Technology Park Corporation, with the objective of identifying Machine learning model to solve the current problems faced by the medical sector in Hong Kong. The system is also now being adopted and used by one of the major public hospitals in Hong Kong.
"With more hospitals adopt our system, we can accumulate more data and further enhance the accuracy to almost perfect level, said Calvin, "this is also the merits of AI and the Internet of Things (IoTs) technology that everybody can get benefits by sharing data to big data analytics for system improvement. We look forward to the system being employed in the private hospitals and other small clinics where small surgeries are allowed to take place with even less manpower for support".
Blueinno's team winning the AI Challenges with the system
Vision: Building the smart hospital
"Looking ahead, the system is only the first step of our smart hospital project", said Monica, "Blueinno AI is specializing in AI, machine learning, and computer vision. And we believe these three technologies are critical to developing the smart hospital".
The above technologies can have 6 major applications for the smart hospital, which is defined as " a hospital that relies on optimised and automated processes built on an ICT environment of interconnected assets, particularly based on Internet of things (IoT), to improve existing patient care procedures and introduce new capabilities" (European Union Agency For Network And Information Security, 2016).
The six applications are:
- 1. Assistant for clinicians in consultation and diagnosis
- 2. Auto reporting for patients
- 3. Monitoring the vital signs of patients
- 4. Imaging and testing
- 5. Detection of abnormal behaviors, such as falling
- 6. Counting of surgical instruments
The concept of the smart hospitals
Educating the young and consulting industries
On the other hand, Blueinno also recognize the importance of educating the young and passing down the technologies to the public so as to encourage everyone to innovate and advancing society together. Therefore, they have developed various AI and machine learning courses for students of all ages, from 10 years old to learn foundational skills such as python and YOLO, to the adult students to learn about deep learning in the programme incorporated with Nvidia.
Apart from individual courses and programmes, Blueinno AI also took up consultancy service in collaboration with other companies and NGOs, such as the MakerBay Foundation on development of AI coral scanning system to help preservation of coral in the Hong Kong water.
Blueinno AI's team provide consultancy service and training to the collaborating organization