AI'S Usage in Healthcare: Assisting Clinical Diagnoses
According to Americans, artificial intelligence’s (AI) usage in healthcare seems untrustworthy. Sixty percent of American adults reported that they would feel uncomfortable if their health care provider relied on AI, according to a 2022 survey conducted by the Pew Research Center. There are various legal concerns with liability that AI technology needs to get through to be fully functioning by itself. However, the machine learning aspect of AI is already being clinically studied for its benefit in helping determine diagnoses.
Machine learning is a field of AI that learns from data to produce and predict relationships regarding the population of the data, according to UCLA Extension. A study by Athreya, Melehy, Suthahar, et al looked into how machine learning could improve the diagnosis of whether a thyroid lump, medically known as a nodule, is cancerous or not (malignant or benign, respectively). The current technique for determining whether a nodule is malignant or benign is commonly done by running a molecular testing (MT) through sequencing the cell samples collected from the nodule, called fine needle aspiration (FNA). This method has a high success rate in ruling out malignancy of the cells, but it lacks in specificity and positive predictive value (PPV), which is determining whether the person who received a positive test actually has malignant cells or not. This is important because a lower PPV means that there is a higher chance of patients who have a benign nodule being given a positive result, which could lead to those misdiagnosed patients receiving unnecessary surgery to remove the benign nodule.
Machine learning is a field of AI that learns from data to produce and predict relationships regarding the population of the data, according to UCLA Extension. A study by Athreya, Melehy, Suthahar, et al looked into how machine learning could improve the diagnosis of whether a thyroid lump, medically known as a nodule, is cancerous or not (malignant or benign, respectively). The current technique for determining whether a nodule is malignant or benign is commonly done by running a molecular testing (MT) through sequencing the cell samples collected from the nodule, called fine needle aspiration (FNA). This method has a high success rate in ruling out malignancy of the cells, but it lacks in specificity and positive predictive value (PPV), which is determining whether the person who received a positive test actually has malignant cells or not. This is important because a lower PPV means that there is a higher chance of patients who have a benign nodule being given a positive result, which could lead to those misdiagnosed patients receiving unnecessary surgery to remove the benign nodule.
Machine learning can utilize healthcare data and analyze the results that could help diagnose potentially cancerous thyroid nodules.
Image Source: Pexels
Therefore, in hopes of improving the specificity and PPV, the researchers utilized machine learning to analyze data incorporating ultrasound imaging data and MT. This was tested retrospectively, meaning the study utilized already existing data to verify the accuracy of AI, using the data of 333 patients with indeterminate thyroid nodules. Multiple analysis was run, and the machine learning’s results were evaluated by a five fold cross validation, a method that validates each trial of the model. With information about the nodule from both imaging and molecular aspects, machine learning demonstrated improved performance than just MT itself. These results were proven to be statistically significant by Athreya, Melehy, Suthahar, et al.
This means that AI helped improve the accuracy of determining whether a thyroid nodule was cancerous, or in other words, malignant. By having AI as an assistant that increases the performance of the diagnostic tests, AI can reduce the number of false positives - patients who unnecessarily surgically remove benign thyroid nodules. However, this study only looked into a single dataset of patients from the UCLA Medical Center, so the results cannot be generalized for the entire population.
With technology such as AI growing rapidly in the modern world, finding ways that can benefit from the technology without threatening the patient’s trust in healthcare is important. Through utilizing AI and machine learning to assist in the diagnosis of potentially cancerous nodules, it provides a positive perspective on how AI can be utilized as a tool in healthcare to aid patient welfare.
This means that AI helped improve the accuracy of determining whether a thyroid nodule was cancerous, or in other words, malignant. By having AI as an assistant that increases the performance of the diagnostic tests, AI can reduce the number of false positives - patients who unnecessarily surgically remove benign thyroid nodules. However, this study only looked into a single dataset of patients from the UCLA Medical Center, so the results cannot be generalized for the entire population.
With technology such as AI growing rapidly in the modern world, finding ways that can benefit from the technology without threatening the patient’s trust in healthcare is important. Through utilizing AI and machine learning to assist in the diagnosis of potentially cancerous nodules, it provides a positive perspective on how AI can be utilized as a tool in healthcare to aid patient welfare.
Featured Image Source: StockSnap
RELATED ARTICLES
|
Vertical Divider
|
Vertical Divider
|
Vertical Divider
|