Artificial intelligence (AI) involves using computerized algorithms to understand and analyze complicated data. Clinically, AI used for diagnostic imaging covers a wide variety of clinical conditions. These tools have shown incredible accuracy, sensitivity, and specificity for detecting small radiographic abnormalities. These tools show great potential to improve public health. Unfortunately, AI imaging studies are limited to only detecting the lesions . These tools tend to overlook the type and biological aggressiveness of a lesion, which creates a skewed representation of AI’s performance.
An artificial intelligence tool, ChestLink reads chest X-rays without oversight from a radiologist, got regulatory clearance in the European Union in April, 2022. The tool scans chest X-rays and automatically sends reports to the patients if there are no abnormalities. Any scans that have pathologies or abbreviations are sent to senior radiologists for assessment.
Why was artificial intelligence introduced in radiology?
- With an increasing patient load and the skewed doctor to patient ratio, it has become empirical to reduce the burden on healthcare services. Automation and AI tools help to ease the stress on radiologists.
- Errors in diagnostic radiology are common. However, there are various preventive mechanisms within the medical system to avoid such errors such as:
- 1. multidisciplinary meetings,
- 2. second opinions,
- 3. subspecialty expertise,
- 4. clinician experience.
What are the methods of AI in medical imaging?
- The first method uses features extracted from regions of interest on the basis of expert knowledge. Examples of these features in cancer characterization include tumor volume, shape, texture, intensity, and location. The most robust features are selected and fed into machine learning classifiers. A large volume of data is sent into the system for assessment and analysis.
- The second method uses deep learning which comprises of multiple layers where feature extraction, selection, and ultimate classification are performed simultaneously during training.
What are the benefits of AI in Diagnostic Radiology?
- Triage: AI screens examinations for the probability of disease and determines the order of interpretation.
- Replacement: The tool is used where its results outperform human radiologists, especially in the case of Chest X-ray assessment.
- Add-on: AI supports existing clinical pathways by handling time-intensive activities. It helps in choosing a personalized patient protocol and estimating the radiation risk. It can also be used for standardizing reports.
What are the drawbacks of using AI?
- Limited Scope: AI is limited in its ability to only focus on a few particular findings. Compared to human radiologists who look at the image as a whole while considering multiple pathologies simultaneously.
- Selection Bias: The data curation technique based on “small focused datasets” has its inherent flaws of selection bias. Researchers may purposefully select the more obvious cases of the pathology to inflate its numbers during testing.
- Flaws in research: To assess the positive imaging initially used to train the software, the expertise, and experience of the radiologist are completely at the will of the researcher. These researchers are less likely to approach senior radiologists for the same.
- Testing Bias: The comparative results as of the AI tool are reported without revealing the expertise of the “average radiologist” used in comparison.