Surgical Technology and Natural Language Processing
Surgical Technology presents excellent challenges that could be simplified with machine learning. The Certified Surgical Technology(CST) practice exam from gotestprep.com contains 176 questions on surgical best practices, and machine learning can help determine the frequency of topics that will appear on the practice CST exam.
Natural Language Processing is the use of Data Science to organize messy language data. Data Science is used to scrape, clean, and processes questions and answers from gotestprep.com in order to organize a list of surgical terms that will appear most frequently.
Surgical Technology is the domain knowledge of sterile and aseptic surgical techniques used to prepare equipment for surgery. Surgical Technologists must be licensed by the Association of Surgical Technologists (AST) in order to work at a hospital.
Check out the code of this experiment at Github.
Audience
This article is largely technical as it is directed at two different peer groups, Surgical Technologists and Data Scientists. Surgical Technologists who need to take or retake the CST exam may find value in these results and should concentrate on questions that will appear most frequently. Data Scientists may also find value in these NLP techniques and it’s implementation to solve a real world problem.
Results
The following pyldavis visualization targets the ‘saliency’ of terms, which is:
To view the codepen, click on the ‘Run Pen’ button below
Explore the interactive model by clicking on the circles under the Intertopic Distance Map
The Codepen visualization is best viewed at 0.5x size on desktop and .25X on mobile.
The conjecture of the above visualization could be that the majority of the questions of the CST exam are about Tissue Removal and Steam Sterilization. These Can be funneled into a Vector Representation of our dataset to find the answers to common questions about Tissue Removal and Steam Sterilization.
Tissue Removal
The goal of most non-elective surgeries is to remove cancerous or infected tissues.
Dilating drops are called mydriatics and are used to dilate the pupil for examination of the retina, refraction testing, or easier removal of the lens.
The tools used in Cataract surgeries are small, specialized forceps. Some common tools used are Bonn Forceps, Utrata Forcepts, and Iris Forceps.
The removal of a tooth or teeth is an extraction procedure. The resection of the soft tissue and excision of the bone surrounding the tooth prior to the removal is called odontectomy.
Odontectomies are dental procedures and are usually used to treat tooth decay. Specialized Dental Extraction Forceps are used during this operation.
McBurney incision is primarily used for removal of the appendix.
Appendectomies are common abdominal procedures used to remove the appendix, and McBurney’s incisions consist of cutting into muscle for appendectomies. A McBurney Retractor is often used to open the abdomen during these surgeries.
Steam Sterilization
Other most frequent terms are about Steam Sterilization, which is newer technology that sterilizes equipment faster but at a much lower temperature. Because of it’s lower temperatures, Steam Sterilization requires more attention to do correctly compared to the classical dry sterilization method.
The four factors of steam sterilization are pressure, temperature, moisture and time. Pressure is used to increase the temperature of the steam to the level where it will kill microbes including spores.
Steam sterilizers are newer technology and are used for heat sensitive equipment like silicone because it sterilizes at a lower temperature. Steam sterilizers are also immediately available in the Operating Room just in case vital surgical instruments get dropped.
The steam sterilization biological indicator must be incubated for 24 hours before the reading is recorded.
Biological indicators are saved and then cataloged in case of infections and malpractice suits. BI’s are integral for best sterility practices and to the patient’s safety.
The biological indicator (BI) contains Bacillus stearothermophilus which is killed when exposed to steam sterilization conditions. Therefore, the BI is the only test that guarantees sterility.
NLP Procedure
The following are Data Science tools used to create our data results.
Data Scraping with Scrapy
Data scraping is used to crawl the internet and extract data. Scrapy is the quickest way to build a web scraper when done correctly.
More information on how to use Scrapy can be found here .
Data Cleaning with BeautifulSoup
BeautifulSoup is often used for data scraping, but it also excels at data cleaning. Scrapy is quick and efficient but it is also messy when scraping HTML tags. BeautifulSoup’s html parser can be used to remove html tags. The scraped data can be transformed from this:
'<div class="su-spoiler-content su-u-clearfix su-u-trim">\xa0<strong><span style="color: #008000;">Correct option: 2<br>\n</span></strong>Sol: The cochlea is a bony spiral canal in the ear.\n</div>',Into this:
'The cochlea is a bony spiral canal in the ear.',Topic Modeling with pyLDAvis
In machine learning and Natural Language Processing, a topic model is used to discover how many abstract topics exist in a collection of documents. The topics of steam sterilization, and tissue removal occurs the most in this experiment while using pyLDAvis.
Vector Representation with sklearn
Vector Representation as a search algorithm can be used to determine how close these questions are related to one another with the use of eigenvectors. The topics of steam sterilization and tissue removal from pyLDAvis can be inputted into the Vector Representation model in order to search for related questions and answers.
Here are the following methods used to create the Vector Representation:
In Conclusion
Surgical Technology presents excellent real world challenges that could be solved with machine learning, and the conjecture of the topic model shows that that the CST practice exam is mostly about Tissue Removal and Steam Sterilizers.
The Data Science tools, Scrapy for data scaping, BeautifulSoup for data cleaning, sklearn for vector representation and pyldavis for topic modeling all can be used to give valuable insights on what to expect on the CST exam.
