. 2021 Sep 2;28(38):52810–52831. doi: 10.1007/s11356-021-16223-0

Table 3.

Role of artificial intelligence in different cancers

Cancer	Background	Internal & external validation	Finding	Reference
Breast cancer	Evaluation of an AI system for breast cancer screening in which they introduce an artificial intelligence	The AI system outperformed all of the human readers in their analysis of six radiologists: the AI system’s area under the receiver operating characteristic curve (AUC-ROC) was 11.5%>normal radiologist’s AUC-ROC.	Clinical studies to increase the precision and reliability of breast cancer screening are now possible thanks to this thorough evaluation of the AI scheme.	(McKinney et al. 2020)
Prostate cancer	A blinded clinical validation analysis and implementation of an artificial intelligence (AI)-based algorithm in a pathology lab for routine clinical use to assist prostate diagnosis was described.	Set of 32 prostate CNB cases (selected from cases occurring between August 2014 and January 2018), comprising 159 parts, to calibrate the algorithm for UPMC-specific whole-slide image attributes (e.g., scanner and staining) and to verify the technical validity of the whole-slide images (e.g., file format and resolution). Internal test:benign vs cancer AUC was 0·997, specificity: 90·14% sentivity to be found	An AI-based algorithm for accurately detecting, grading, and evaluating clinically relevant findings in digitized slides of prostate CNBs was developed, externally clinically validated, and deployed in clinical practice.	(Pantanowitz et al. 2020)
Lung cancer	End-to-end design is particularly important when considering AI application in lung cancer	In the test dataset consisting of 6716 cases (86 cancer-positives) from the National Lung Screening Trial (NLST), this model achieved an AUC of 94.4% for lung cancer-risk prediction, which is considered to be a state-of-the-art performance. The model performed similarly with an AUC of 95.5% on an independent clinical validation set of 1139 cases.	Demonstrating a step toward automated image evaluation for lung cancer risk estimation using AI.	(Ardila et al. 2019)

Cancer

Background

Internal & external validation

Finding

Reference

Breast cancer

Evaluation of an AI system for breast cancer screening in which they introduce an artificial intelligence

The AI system outperformed all of the human readers in their analysis of six radiologists: the AI system’s area under the receiver operating characteristic curve (AUC-ROC) was 11.5%>normal radiologist’s AUC-ROC.

Clinical studies to increase the precision and reliability of breast cancer screening are now possible thanks to this thorough evaluation of the AI scheme.

(McKinney et al. 2020)

Prostate cancer

A blinded clinical validation analysis and implementation of an artificial intelligence (AI)-based algorithm in a pathology lab for routine clinical use to assist prostate diagnosis was described.

Set of 32 prostate CNB cases (selected from cases occurring between August 2014 and January 2018), comprising 159 parts, to calibrate the algorithm for UPMC-specific whole-slide image attributes (e.g., scanner and staining) and to verify the technical validity of the whole-slide images (e.g., file format and resolution). Internal test:benign vs cancer

AUC was 0·997, specificity: 90·14% sentivity to be found

An AI-based algorithm for accurately detecting, grading, and evaluating clinically relevant findings in digitized slides of prostate CNBs was developed, externally clinically validated, and deployed in clinical practice.

(Pantanowitz et al. 2020)

Lung cancer

End-to-end design is particularly important when considering AI application in lung cancer

In the test dataset consisting of 6716 cases (86 cancer-positives) from the National Lung Screening Trial (NLST), this model achieved an AUC of 94.4% for lung cancer-risk prediction, which is considered to be a state-of-the-art performance. The model performed similarly with an AUC of 95.5% on an independent clinical validation set of 1139 cases.

Demonstrating a step toward automated image evaluation for lung cancer risk estimation using AI.

(Ardila et al. 2019)