InferRead AI: Empowering Early-Career Radiologists with Clinical-Grade Imaging Assistance

Currently, the application of AI in healthcare is becoming increasingly widespread. In particular, AI-assisted diagnostic systems for medical imaging have already covered screening for various common types of cancer, and these products can now be seamlessly integrated into physicians’ clinical workflows. During the pilot implementation of medical AI products, what kind of support will physicians receive, and who stands to benefit the most from the deployment of AI?. In this regard, VCBeat recently gathered insights from some physicians at the First National Radiology Residency Training Physicians’ Imaging Skills Competition.

At the inaugural National Radiology Residency Imaging Skills Competition, Infervision provided the designated AI-assisted screening product for the event, realistically simulating the entire clinical workflow of radiologists. During the competition, Infervision’s AI assisted participants in the image interpretation segment, enabling physicians who had just begun their frontline work in radiology to successfully complete high-difficulty tasks. These tasks included the detection of tiny nodules, identification of ground-glass nodules, and differentiation of nodules on thin-slice images.

Furthermore, it is worth noting that during the semi-final and final rounds, Infervision’s products fully demonstrated their user-friendly features. Contestants who were using Infervision’s products for the first time were able to master them after only a brief training session of a few minutes prior to the competition.

VCBeat learned from Infervision that most of the doctors participating in this competition are radiologists working on the front lines. The integration of AI into clinical practice has provided the greatest benefit to physicians who have recently begun their frontline work.

In the radiology department, the department head and several renowned radiologists carry heavy workloads, much of which involves research and teaching. Only complex or challenging imaging cases are referred to these experts for final diagnosis. Some general physicians also seek their expertise when uncertain about certain images, but such cases constitute only a portion of the hospital’s total patient volume.

According to VCBeat, chief and associate chief physicians in the radiology departments of many large tertiary hospitals have limited weekly working hours. They rely on general physicians to perform initial screenings of the hundreds of imaging studies reviewed daily, with these senior specialists stepping in for consultations only on complex or difficult cases.

For radiologists whose primary responsibility is image interpretation, particularly those newly entering frontline clinical practice, insufficient work experience combined with a heavy workload can lead to unnecessary missed diagnoses.

Upon investigation, the primary concern for physicians is missed diagnoses. Medical malpractice incidents arising from missed diagnoses can have severely detrimental effects on the career trajectories of radiologists. In addition to missed diagnoses, both radiologists and clinicians are equally concerned about excessive false positives. An high rate of false positives not only increases physicians’ workload and leads to overtreatment, but also induces anxiety in patients, thereby adversely affecting subsequent monitoring, follow-up, and treatment.

In the diagnostic workflow enhanced by inferential AI-assisted diagnosis, physicians require only brief training and adaptation. Leveraging the intelligent pre-processing capabilities of AI products, all predictive results are instantly displayed with a single click. Physicians need only eliminate a small number of false positives, thereby easily reducing the time required for pulmonary nodule interpretation to just over ten seconds.

Additionally,The Use of Medical AI Products: A Mutual Learning Process for Junior Frontline Radiologists Newly Entering Clinical PracticeAfter several years of refinement and iteration, Infervision’s medical AI products have gained recognition. In daily practice, general radiologists often refer to the recommendations provided by medical AI, which also serves as a process for their own learning and professional development. Meanwhile, as medical AI systems are not yet perfect, physicians’ feedback helps drive the iterative improvement of these AI products.

For the department, with the assistance of medical AI, physicians can achieve rapid professional growth, shorten training periods, and alleviate the teaching burden on specialist physicians.

The greatest additional benefit that medical AI products bring to frontline clinicians upon clinical adoption is liberating them from the heavy burden of image interpretation, thereby freeing up time for clinical research.。

Professor Liu Shiyuan, Director of the Department of Diagnostic and Interventional Radiology and Nuclear Medicine at Changzheng Hospital, stated at the 5th Qianjiang International Medical Imaging Forum that image interpretation is the most time-consuming task for radiologists and, at the same time, the task most suitable for delegation to AI. Only by freeing radiologists from image interpretation can they dedicate time to clinical research and achieve rapid professional growth.

Medical practice is a rigorous endeavor. For radiologists working on the front lines, their confidence in using Infervision’s products stems primarily from the robustness, ease of use, and accuracy that have been recognized by expert physicians.

Regarding usability, as mentioned above, all prediction results are displayed instantly with a single click by the physician. The physician’s task is merely to eliminate a small number of false positives.

Regarding robustness and accuracy, Infervision stated that its AI medical products deliver reliable analytical results for CT scanners from various manufacturers, covering both standard-dose and low-dose screening protocols. The system demonstrates an extremely low missed diagnosis rate and a low false-positive rate. In particular, it exhibits high sensitivity to small nodules and ground-glass opacity (GGO) nodules, effectively helping physicians reduce missed diagnoses that are prone to occur under high-intensity workloads.

Dr. Xie from the First Hospital of Fuzhou, affiliated with Fujian Medical University, commented: “Currently, our product demonstrates high sensitivity in AI-based nodule detection. It can effectively screen for proliferative lesions and tiny nodules, achieving robust performance with minimal missed cases. Further targeted differential diagnosis is still required, but the overall workflow has become significantly easier compared to before.”

Infervision stated that, as of February 2018, its AI solutions had been deployed and gone live in seven of the top 10 hospitals on Fudan University’s national hospital ranking list, and had entered 25 of the top 50 hospitals.

Soon, the company will launch version 4.0 of its product, which will assist in the diagnosis and mitigation of condition 14. In the future, Infervision will not only serve the radiology departments of tertiary hospitals but also extend its services to grassroots medical institutions with scarcer healthcare resources, meeting the growing demand for imaging screening.