| 1. |
|
|
| 2. |
- Olsson, Henrik, et al.
(författare)
-
Estimating diagnostic uncertainty in artificial intelligence assisted pathology using conformal prediction
- 2022
-
Ingår i: Nature Communications. - : Springer Nature. - 2041-1723. ; 13:1
-
Tidskriftsartikel (refereegranskat)abstract
- Unreliable predictions can occur when an artificial intelligence (AI) system is presented with data it has not been exposed to during training. We demonstrate the use of conformal prediction to detect unreliable predictions, using histopathological diagnosis and grading of prostate biopsies as example. We digitized 7788 prostate biopsies from 1192 men in the STHLM3 diagnostic study, used for training, and 3059 biopsies from 676 men used for testing. With conformal prediction, 1 in 794 (0.1%) predictions is incorrect for cancer diagnosis (compared to 14 errors [2%] without conformal prediction) while 175 (22%) of the predictions are flagged as unreliable when the AI-system is presented with new data from the same lab and scanner that it was trained on. Conformal prediction could with small samples (N = 49 for external scanner, N = 10 for external lab and scanner, and N = 12 for external lab, scanner and pathology assessment) detect systematic differences in external data leading to worse predictive performance. The AI-system with conformal prediction commits 3 (2%) errors for cancer detection in cases of atypical prostate tissue compared to 44 (25%) without conformal prediction, while the system flags 143 (80%) unreliable predictions. We conclude that conformal prediction can increase patient safety of AI-systems.
|
|
| 3. |
- van der Kwast, Theo, et al.
(författare)
-
International Society of Urological Pathology Expert Opinion on Grading of Urothelial Carcinoma
- 2022
-
Ingår i: European Urology Focus. - : Elsevier BV. - 2405-4569. ; 8:2, s. 438-446
-
Forskningsöversikt (refereegranskat)abstract
- Context: Grading is the mainstay for treatment decisions for patients with non–muscle-invasive bladder cancer (NMIBC). Objective: To determine the requirements for an optimal grading system for NMIBC via expert opinion. Evidence acquisition: A multidisciplinary working group established by the International Society of Urological Pathology reviewed available clinical, histopathological, and molecular evidence for an optimal grading system for bladder cancer. Evidence synthesis: Bladder cancer grading is a continuum and five different grading systems based on historical grounds could be envisaged. Splitting of the World Health Organization (WHO) 2004 low-grade class for NMIBC lacks diagnostic reproducibility and molecular-genetic support, while showing little difference in progression rate. Subdividing the clinically heterogeneous WHO 2004 high-grade class for NMIBC into intermediate and high risk categories using the WHO 1973 grading is supported by both clinical and molecular-genetic findings. Grading criteria for the WHO 1973 scheme were detailed on the basis of literature findings and expert opinion. Conclusions: Splitting of the WHO 2004 high-grade category into WHO 1973 grade 2 and 3 subsets is recommended. Provision of more detailed histological criteria for the WHO 1973 grading might facilitate the general acceptance of a hybrid four-tiered grading system or—as a preferred option—a more reproducible three-tiered system distinguishing low-, intermediate (high)-, and high-grade NMIBC. Patient summary: Improvement of the current systems for grading bladder cancer may result in better informed treatment decisions for patients with bladder cancer. A three-tiered grading system for non–muscle invasive bladder cancer derived by splitting the heterogeneous World Health Organization (WHO) 2004 high-grade category into WHO 1973 grade 2 and 3 subsets is recommended, as this may result in more informed treatment decisions.
|
|
