Illinois: According to a study published in Radiology, a journal of the Radiological Society of North America (RSNA), the attendance of a artificial intelligence Improved (AI) algorithm with high diagnostic accuracy radiologist performance in the detection of lung cancers on chest X-rays and greater human acceptance of AI suggestions.
Although AI-based imaging has advanced rapidly in the field of medicine, the factors that affect radiologists’ diagnostic determinations in AI-assisted image reading remain unexplored.
Researchers from Seoul National University looked at how these factors might influence the detection of malignant pulmonary nodules during AI-assisted reading of chest X-rays.
In this retrospective study, 30 readers, including 20 chest radiologists with 5 to 18 years of experience and 10 radiology residents with only 2 to 3 years of experience, evaluated 120 non-AI chest radiographs. Of the 120 chest radiographs evaluated, 60 were from lung cancer patients (32 men) and 60 were controls (36 men). The patients had a median age of 67 years. In a second session, each group reinterpreted the X-rays, with the help of a high-precision or low-precision AI. Readers were blind to the fact that two different AIs were used.
use of the high accuracy The AI improved the detection performance of the readers to a greater extent than the low precision AI. The use of high-precision AI also led to more frequent changes in readers’ determinations, a concept known as susceptibility.
“It is possible that the relatively large sample size in this study has bolstered readers’ confidence in AI suggestions,” said study lead author Chang Min Park, MD, Ph.D., of the Department of Radiology and the Institute of Radiation Medicine in Seoul. Seoul National University College of Medicine. “We think this issue of human trust in AI is what we observed in susceptibility in this study: humans are more susceptible to AI when using high diagnostic throughput AI.”
Compared with the first reading session, High Diagnostic Accuracy AI-assisted readers in the second reading session showed higher sensitivity per lesion (0.63 vs 0.53) and specificity (0.94 vs 0 .88). Alternatively, readers assisted by the low diagnostic accuracy AI at the second reading session showed no improvement between the two reading sessions for any of these measures.
“Our study suggests that AI can help radiologists, but only when the diagnostic performance of AI matches or exceeds that of the human reader,” said Dr. Park.
The results underline the importance of using AI with high diagnostic performance. However, Dr. Park noted that the definition of “high diagnostic throughput AI” may vary depending on the task and the clinical context in which it will be used. For example, an AI model that can detect all abnormalities on chest X-rays may seem ideal. But in practice, such a model would be of limited value in reducing the workload in a mass screening setting for pulmonary tuberculosis.
“Therefore, our study suggests that the clinically appropriate use of AI requires both the development of high-performance AI models for given tasks and considerations of the relevant clinical setting to which that AI will be applied,” said Dr. Park. .
In the future, the researchers want to expand their work on human-AI collaboration to other abnormalities in chest X-rays and CT images.
