New study shows combined power of deep learning and radiologists in prostate cancer detection

A new study published in the Journal of Radiology reveals promising advancements in the use of artificial intelligence (AI) to improve radiology reporting, particularly in the context of prostate cancer detection. The research demonstrates that AI can significantly enhance the accuracy and efficiency of radiological assessments, especially in complex cases involving large sets of imaging data.

In a retrospective study involving 5,215 patients and 5,735 multiparametric MRI examinations for prostate cancer evaluation, the performance of a deep learning (DL) model in detecting clinically significant prostate cancer (csPCa) was found to be comparable to that of experienced radiologists. The area under the receiver operating characteristic curve (AUC) for the DL model was 0.86, compared to 0.84 for radiologists (p=0.68), indicating no significant difference in diagnostic accuracy between the two.

However, the study found that combining the DL model with radiologists significantly improved diagnostic performance on an external test set, with an AUC of 0.89 compared to 0.84 for radiologists alone (p<0.001). This combined approach demonstrated the potential of AI to enhance radiologists’ capabilities, leading to more accurate diagnoses.

Furthermore, in cases with positive examinations, gradient-weighted class activation maps consistently highlighted the csPCa lesion, providing visual explanations of the AI’s decision-making process and offering additional insights to radiologists.

These findings underscore the potential of AI to transform radiology practices by providing more efficient and accurate diagnostic tools, ultimately benefiting both healthcare providers and patients. The research highlights the importance of a collaborative approach, where AI augments radiologists’ expertise rather than replacing it.

Read the full article here

Disclosures: This article was created by the touchONCOLOGY team utilizing AI as an editorial tool (ChatGPT (GPT-4o) [Large language model]. https://chat.openai.com/chat.) The content was developed and edited by human editors. No funding was received in the publication of this article.

Looking for more content on prostate cancer? Take a look at some of our recent expert interviews:

ARPi vs. SOC and taxanes in mCRPC – ProBio study with Henrik Grönberg
Current advances in prostate cancer treatment options and what we might expect in the next 5 years with Axel Merseburger