Oops, you're using an old version of your browser so some of the features on this page may not be displaying properly.

MINIMAL Requirements: Google Chrome 24+Mozilla Firefox 20+Internet Explorer 11Opera 15–18Apple Safari 7SeaMonkey 2.15-2.23

Ability of Deep Learning to Accurately Diagnose Frequent STS Subtypes from Conventional Histopathological Slides

When aided by a deep learning model, pathologists make faster and more accurate diagnoses of soft tissue sarcomas
26 Aug 2021
Pathology/Molecular Biology;  Cancer Intelligence (eHealth, Telehealth Technology, BIG Data)
Soft Tissue Sarcomas

Findings from one of the first studies to investigate the use of artificial intelligence in the pathological management of soft tissue sarcoma (STS) show that a deep learning model was able to classify five of the most common STS subtypes from histology alone. When aided by the deep learning model, pathologists were more accurate, faster, and more certain in their diagnosis. A similar deep learning model was able to predict the disease-specific survival status in the most common STS subtype. The deep learning model's prediction was an independent prognostic factor. Furthermore, new image features associated with survival could be identified. The study findings are published by Dr. Sebastian Foersch of the Institute of Pathology, University Medical Center Mainz in Mainz, Germany and colleagues in the Annals of Oncology.

The authors explained in the study background that reproducibility of STS diagnosis is poor across pathologists who are not seeing these tumours regularly and there is a high intra- and inter-observer variability among pathologists with lack of expertise in sarcomas. High level of diagnostic expertise is often limited to a few reference centres. This might contribute to substantially delayed diagnosis which has a significant negative impact on STS outcomes.

The study team used digital pathology and deep learning for diagnosis and prognosis prediction of STS. Their retrospective, multicentre study included 506 histopathological slides from 291 patients with STS. The Cancer Genome Atlas cohort of 240 patients served as training and validation set. A second, multicentre cohort of 51 patients served as an additional test set. The use of the deep learning model as a clinical decision support system was evaluated by 9 pathologists with different levels of expertise. For prediction of prognosis, 139 slides from 85 patients with leiomyosarcoma were used. Area under the receiver operating characteristic (AUROC) and accuracy served as main outcome measures.

The deep learning model achieved a mean AUROC of 0.97 (±0.01) and an accuracy of 79.9% (±6.1%) in diagnosing the five most common STS subtypes. The deep learning model significantly improved the accuracy of the pathologists from 46.3% (±15.5%) to 87.1% (±11.1%). Furthermore, they were significantly faster and more certain in their diagnosis.

In leiomyosarcoma, the mean AUROC in predicting the disease-specific survival status was 0.91 (±0.1) and the accuracy was 88.9% (±9.9%).

Cox regression showed the deep learning model’s prediction is a significant independent prognostic factor (hazard ratio 5.5, 95% confidence interval 1.56-19.7; p = 0.008) in these patients, outperforming other risk factors.

In this study, the pathologists utilised state-of-the-art deep learning approaches to correctly diagnose most common subtypes of STS. They compared their best model's performance with the performance of pathology experts with varying experience in general pathology. They investigated leiomyosarcomas as the most common subtype of STS more thoroughly and were able to differentiate between patients with a good or bad prognosis based on their tumour's histomorphology alone by using artificial intelligence. They explored different visualisation techniques to identify recurring microscopic features associated with different prognosis. The authors concluded that the adoption of artificial intelligence in the pathological management of STS could substantially improve the clinical management of patients with STS.

This work was supported by the German Federal Ministry of Education and Research grant, the University Medical Center Mainz, the Mainz Research School of Translational Biomedicine, and the Manfred-Stolte-Foundation.

Reference

Foersch S, Eckstein M, Wagner D-C, et al. Deep learning for diagnosis and survival prediction in soft tissue sarcoma. Annals of Oncology 2021;32(9):1178-1187. DOI: https://doi.org/10.1016/j.annonc.2021.06.007

This site uses cookies. Some of these cookies are essential, while others help us improve your experience by providing insights into how the site is being used.

For more detailed information on the cookies we use, please check our Privacy Policy.

Customise settings
  • Necessary cookies enable core functionality. The website cannot function properly without these cookies, and you can only disable them by changing your browser preferences.