AI accuracy matches radiologists in detecting gall bladder cancer: study

Hyderabad: A groundbreaking study conducted at the Postgraduate Institute of Medical Education and Research (PGIMER) in Chandigarh, in collaboration with the Indian Institute of Technology (IIT), New Delhi, has showcased the potential of Artificial Intelligence (AI) in revolutionising the detection of gall bladder cancer (GBC).

Published in The Lancet Regional Health Southeast Asia journal, the study reveals that an AI-based approach achieved diagnostic performance comparable to experienced radiologists, offering new hope in the fight against this highly aggressive malignancy.

Gall bladder cancer is a particularly challenging condition to diagnose early due to its poor detection rate and high mortality. Benign gall bladder lesions often exhibit similar imaging features, making early diagnosis elusive.

However, the research team in Chandigarh set out to change this by leveraging Deep Learning (DL) techniques within the realm of AI. Deep Learning, a subset of AI, emulates the human brain’s ability to process data and make decisions.

In this study, abdominal ultrasound data from patients with gall bladder lesions were used, with a DL model being trained on a dataset of 233 patients, validated on 59 patients, and tested on 273 patients.

Deep Learning model provides high accuracy

The DL model’s diagnostic performance was evaluated based on sensitivity, specificity, and the area under the receiver operating characteristic curve (AUC), a widely accepted measure of diagnostic accuracy. Simultaneously, two radiologists independently reviewed the ultrasound images to compare their performance against the DL model.

In the test set, the DL model demonstrated an impressive sensitivity of 92.3 per cent, a specificity of 74.4 per cent, and an AUC of 0.887 for detecting GBC. These results were on par with the performance of the radiologists, highlighting the AI model’s potential.

Moreover, the DL-based approach exhibited high sensitivity and AUC even in challenging scenarios, such as cases involving gallstones, contracted gall bladders, small lesion sizes (less than 10 mm), and lesions in the neck of the gall bladder. It also outperformed one of the radiologists in detecting the mural thickening type of GBC, despite a slight reduction in specificity.

The study’s authors noted, “The DL-based approach demonstrated diagnostic performance comparable to experienced radiologists in detecting GBC using ultrasound.” They further emphasised the need for multicentre studies to fully unlock the potential of DL-based GBC diagnosis.

Cancer specialists weigh in

Leading cancer specialists have welcomed this promising development.

Dr Sarah Patel, a renowned oncologist, said, “The potential of AI in early cancer diagnosis is truly exciting. Detecting gall bladder cancer at an advanced stage is often associated with poor outcomes. AI can be a game-changer in identifying cases at an earlier, more treatable stage.”

Dr Rajesh Khanna, a radiologist with extensive experience in cancer imaging, added, “The study’s results are compelling. AI’s ability to consistently deliver high sensitivity in detecting GBC, even in complex cases, could significantly enhance our diagnostic capabilities.”

However, experts also caution that while this study is a significant step forward, further research, especially multicentre studies, is crucial to validate these findings on a larger scale. Additionally, the rapid evolution of AI and diagnostic technologies means that continuous updates and improvements will be necessary.

This study offers a glimpse into the future of cancer diagnosis, where AI-driven solutions may play a pivotal role in detecting malignancies like gallbladder cancer at an earlier and more manageable stage, potentially saving countless lives.