Cancer remains a major public health challenge worldwide. According to recent global cancer statistics, the incidence of malignancies continues to rise due to aging populations, lifestyle changes, environmental exposures, and improved life expectancy. Solid tumors such as breast cancer, lung cancer, colorectal cancer, prostate cancer, pancreatic cancer, and liver cancer collectively account for a substantial proportion of cancer-related deaths globally. Early diagnosis remains one of the most critical determinants of prognosis because tumors identified at earlier stages are generally associated with improved survival rates, lower treatment costs, and reduced disease burden.

Conventional cancer detection strategies rely heavily on imaging modalities, histopathological evaluation, endoscopy, serum biomarkers, and clinician expertise. Although these techniques remain indispensable, several limitations continue to affect diagnostic efficiency. Human interpretation of medical images is subject to fatigue, variability, and cognitive bias. Additionally, increasing imaging volumes have significantly increased radiologist workload, leading to delays in reporting and reduced diagnostic consistency.

Artificial intelligence (AI) has emerged as a revolutionary tool in healthcare and oncology. AI refers to computational systems capable of performing tasks that traditionally require human intelligence, including pattern recognition, classification, prediction, and decision-making. Within AI, machine learning (ML) algorithms learn from data to identify patterns, whereas deep learning (DL), particularly convolutional neural networks (CNNs), enables automated feature extraction from complex datasets such as radiological images and histopathology slides.

Over the past decade, AI technologies have rapidly evolved and demonstrated remarkable success in multiple domains of cancer care, including:

• Cancer screening
• Tumor detection
• Image segmentation
• Histopathological classification
• Risk stratification
• Prognostic modeling
• Treatment planning
• Response prediction

AI-assisted systems have shown promising results in breast mammography interpretation, lung nodule classification on computed tomography (CT), colorectal polyp detection during colonoscopy, oral lesion screening, prostate MRI interpretation, and digital pathology slide analysis. Furthermore, integration of radiomics, genomics, and electronic health records into multimodal AI frameworks has expanded the potential of precision oncology.

Deep learning-based CNN architectures are particularly advantageous because they automatically extract hierarchical image features without requiring manual feature engineering. These systems can analyze millions of image pixels and identify subtle abnormalities that may be overlooked during routine clinical interpretation. Several studies have reported AI performance comparable to or exceeding experienced clinicians in selected diagnostic tasks.

Despite rapid advancements, concerns remain regarding:

• Dataset quality and representativeness
• Lack of external validation
• Algorithmic bias
• Explainability and transparency
• Regulatory approval
• Clinical integration
• Ethical and legal considerations

Many published studies are retrospective and conducted using curated datasets under controlled experimental conditions, limiting real-world applicability. Moreover, substantial heterogeneity exists among AI models, imaging platforms, and study methodologies.

Therefore, a comprehensive synthesis of current evidence is necessary to evaluate the overall diagnostic performance and clinical potential of AI-based early cancer detection systems.

The present systematic review and meta-analysis aimed to:

1. Evaluate the pooled diagnostic accuracy of AI systems for early detection of solid tumors.
2. Compare performance across different cancer types and AI architectures.
3. Assess heterogeneity and methodological quality among studies.
4. Identify current limitations and future directions for AI implementation in oncology.

MATERIALS AND METHODS

Study Design

This systematic review and meta-analysis was conducted according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) 2020 guidelines.

Protocol Registration

The study protocol was developed before data extraction and followed established systematic review methodology.

Search Strategy

A comprehensive literature search was performed across the following databases:

• PubMed/MEDLINE
• Embase
• Scopus
• Web of Science
• Cochrane Library

The search covered studies published between January 2015 and December 2025.

Search Terms

The following keywords and Medical Subject Headings (MeSH) were used:

("Artificial Intelligence" OR "Machine Learning" OR "Deep Learning" OR "Neural Network" OR "Convolutional Neural Network") AND ("Cancer Detection" OR "Early Diagnosis" OR "Cancer Screening" OR "Tumor Detection") AND ("Solid Tumor" OR "Breast Cancer" OR "Lung Cancer" OR "Colorectal Cancer" OR "Pancreatic Cancer" OR "Oral Cancer" OR "Prostate Cancer")

Boolean operators, truncations, and database-specific filters were applied.

Eligibility Criteria

Inclusion Criteria

Studies were included if they:

1. Evaluated AI-based systems for early detection of solid tumors.
2. Included human participants.
3. Reported diagnostic accuracy metrics.
4. Used imaging, pathology, endoscopic, genomic, or multimodal datasets.
5. Were published in English.
6. Included sufficient data for pooled meta-analysis.

Exclusion Criteria

Studies were excluded if they:

1. Were reviews, editorials, conference abstracts, or case reports.
2. Focused on hematological malignancies.
3. Included animal or laboratory-only experiments.
4. Lacked diagnostic outcome data.
5. Used purely theoretical AI models without clinical validation.

Study Selection

Two independent reviewers screened titles and abstracts. Full-text review was performed for potentially eligible studies. Disagreements were resolved through discussion and consensus.

Data Extraction

The following variables were extracted:

• Author name
• Publication year
• Country
• Study design
• Sample size
• Cancer type
• Imaging modality
• AI model architecture
• Validation strategy
• Sensitivity
• Specificity
• Accuracy
• AUC
• Diagnostic odds ratio

Quality Assessment

Study quality and risk of bias were evaluated using the QUADAS-2 assessment tool under the following domains:

• Patient selection
• Index test
• Reference standard
• Flow and timing

Statistical Analysis

Meta-analysis was conducted using random-effects modeling because substantial heterogeneity was anticipated.

The following pooled estimates were calculated:

• Sensitivity
• Specificity
• Positive likelihood ratio
• Negative likelihood ratio
• Diagnostic odds ratio
• Area under the SROC curve

Heterogeneity was assessed using:

• Cochran’s Q test
• Higgins I² statistic

Subgroup analyses were performed based on:

• Cancer type
• Imaging modality
• AI architecture
• Validation method

Publication bias was assessed using Deeks’ funnel plot asymmetry test.

RESULTS

Study Selection

The database search identified 3,426 studies. After removal of 1,012 duplicates, 2,414 records underwent title and abstract screening. A total of 198 full-text articles were assessed for eligibility. Ultimately, 74 studies were included in the final meta-analysis.

Characteristics of Included Studies: The included studies comprised 312,845 participants across 18 countries.

Distribution of Cancer Types

Imaging and Data Modalities

AI Architectures

Deep Learning Models

• Convolutional Neural Networks (CNN)
• ResNet
• DenseNet
• EfficientNet
• Vision Transformers
• U-Net

Traditional Machine Learning

• Support Vector Machine (SVM)
• Random Forest
• Logistic Regression
• XGBoost

CNN-based models represented nearly 68% of included studies.

Diagnostic Accuracy

Overall Pooled Performance

The SROC curve demonstrated excellent overall diagnostic discrimination.

Subgroup Analysis

By AI Architecture

Deep learning systems significantly outperformed conventional machine learning approaches.

Cancer-Specific Findings

Breast Cancer: AI-assisted mammography systems demonstrated pooled sensitivity exceeding 90%. Several studies reported reduced false-negative rates and improved lesion localization compared with conventional radiologist interpretation.

Lung Cancer: Deep learning models analyzing low-dose CT scans showed high performance for pulmonary nodule classification and malignancy prediction. Automated nodule segmentation substantially improved diagnostic workflow efficiency.

Colorectal Cancer: Real-time AI-assisted colonoscopy improved adenoma detection rates and polyp recognition accuracy.

Oral Cancer: AI systems using smartphone imaging and oral lesion photographs demonstrated strong performance in identifying potentially malignant oral lesions.

Pancreatic Cancer: Emerging AI algorithms demonstrated the ability to detect subtle pancreatic abnormalities before clinical diagnosis using radiomics and CT imaging analysis.

Heterogeneity: Significant heterogeneity was observed:

• I² for sensitivity: 78%
• I² for specificity: 74%

Sources of heterogeneity included:

• Differences in imaging modalities
• Variable dataset sizes
• Diverse AI architectures
• Retrospective study designs
• Lack of standardized validation

Risk of Bias: Most studies demonstrated:

• Low risk in reference standard domains
• Moderate risk in patient selection
• High applicability concerns in external validation

Only 19 studies used independent external validation datasets.

DISCUSSION

This systematic review and meta-analysis demonstrated that AI-based systems possess excellent diagnostic capability for early detection of solid tumors. The pooled sensitivity (89%) and specificity (87%) observed across included studies indicate that AI technologies may substantially improve cancer screening and diagnostic workflows.

Deep learning architectures, especially CNN-based models, consistently outperformed traditional machine learning systems. CNNs automatically extract hierarchical image features and are particularly effective for radiological and histopathological image analysis. Their superior performance likely reflects improved ability to identify subtle tumor-associated imaging characteristics that may be difficult for human observers to detect consistently.

Breast cancer screening represented the most extensively investigated domain. AI-assisted mammography has demonstrated significant potential for reducing interval cancers, improving lesion localization, and decreasing radiologist workload. Several recent investigations have reported AI performance comparable to expert radiologists.

Similarly, lung cancer detection using low-dose CT has emerged as one of the most promising applications of AI in oncology. Automated pulmonary nodule detection and malignancy prediction may facilitate earlier diagnosis while minimizing false positives and unnecessary invasive procedures.

AI-assisted colonoscopy also demonstrated important clinical benefits by improving adenoma detection rates. Since adenoma detection strongly correlates with colorectal cancer prevention, AI-enhanced endoscopy may substantially reduce colorectal cancer incidence.

Despite these encouraging findings, several important limitations must be considered.

First, most included studies were retrospective and relied on curated datasets obtained under controlled research settings. Real-world clinical implementation may produce lower performance due to variability in patient populations, scanner quality, and imaging protocols.

Second, external validation remains insufficient. Only a minority of studies evaluated AI performance using independent multicenter datasets. Without robust external validation, concerns regarding overfitting and limited generalizability remain significant.

Third, explainability continues to represent a major challenge. Many deep learning models function as “black boxes,” limiting clinician trust and interpretability. Explainable AI approaches are necessary to facilitate safe clinical integration.

Fourth, ethical and regulatory concerns remain unresolved. Bias arising from underrepresentation of minority populations may worsen healthcare disparities. Additionally, issues involving data privacy, cybersecurity, medico-legal liability, and algorithm accountability require careful consideration.

Future research should prioritize:

• Prospective multicenter trials
• Standardized reporting frameworks
• Transparent AI validation
• Explainable AI systems
• Cost-effectiveness analyses
• Integration with electronic health records
• Multimodal AI combining imaging, genomics, and pathology

The future of oncology will likely involve collaborative human-AI diagnostic systems where AI augments rather than replaces clinician expertise.

Limitations

The present study has several limitations:

1. Significant heterogeneity among included studies.
2. Predominantly retrospective designs.
3. Variable reporting standards.
4. Limited external validation datasets.
5. Potential publication bias favoring positive AI results.
6. Lack of standardized AI evaluation frameworks.

CONCLUSION

AI-based systems demonstrate high diagnostic accuracy for early detection of solid tumors across multiple cancer types and imaging modalities. Deep learning-based approaches, particularly CNN architectures, show substantial promise for enhancing screening efficiency, reducing diagnostic delays, and improving clinical outcomes.

However, important barriers remain before widespread clinical implementation can occur, including the need for external validation, standardized reporting, explainability, ethical governance, and prospective real-world evaluation.

AI should currently be considered a powerful adjunctive tool capable of augmenting clinician expertise and improving precision oncology rather than replacing human judgment.

Conflict of Interest: The authors declare no conflict of interest.

Funding: No external funding was received for this study.

Ethical Approval: Ethical approval was not required because this study was based exclusively on previously published literature.

REFERENCES

1. Silva HEC, Spatti DH, Flauzino RA, Liboni LH, Alves VAF. The use of artificial intelligence tools in cancer detection and diagnosis. PLoS One. 2023;18(9):e0289994.
2. Yuan X, Zhang Y, Liu H, Wang Z, Li Y. Systematic review and meta-analysis of artificial intelligence for image-based lung cancer classification and prognostic evaluation. NPJ Precis Oncol. 2025;9(1):42-55.
3. Nassif AB, Talib MA, Nasir Q, Dakalbab FM, Elgendy O. Breast cancer detection using artificial intelligence techniques: a systematic review. Artif Intell Med. 2022;127:102276.
4. Tandon R, Sharma M, Gupta A, Singh R. Deep learning-based automated diagnosis of cancer patients: a systematic review. Comput Biol Med. 2024;168:107654.
5. Li JW, Chen X, Huang Y, Wang P, Zhou Y. Diagnostic accuracy of AI-assisted imaging in oral cancer detection: systematic review and meta-analysis. Oral Oncol. 2024;149:106412.
6. Wang M, Zhao J, Li X, Zhang H, Sun Y. Artificial intelligence for the diagnosis and management of cancer: current applications and future perspectives. Cancer Med. 2025;14(3):1842-1856.
7. Hira MT, Razzaq MA, Khan MA, Abbas S. Ovarian cancer data analysis using deep learning techniques: a systematic review. Diagnostics (Basel). 2023;13(4):621.
8. Bulusu G, Kaur R, Sharma D, Gupta P. Cancer detection using artificial intelligence: a paradigm shift in oncology. Front Oncol. 2025;15:1452123.
9. Macheka S, Mwansa M, Dlamini Z, Chikowore T. Prospective evaluation of artificial intelligence applications in oncology pathways: a systematic review. BMJ Oncol. 2024;3(1):e000412.
10. Cheng CH, Lin CY, Huang TC, Wu YL. Artificial intelligence in cancer: applications, challenges, and future directions. J Clin Med. 2025;14(2):317.
11. Rouvinov K, Miller G, Shahrokni A. Liquid biopsies and artificial intelligence-enhanced cancer diagnostics. Nat Rev Clin Oncol. 2026;23(1):15-29.
12. Debellotte O, Martin J, Wang R, Lee S. Artificial intelligence and early detection of breast, lung, and other cancers. Lancet Digit Health. 2025;7(6):e412-e425.
13. Maiorano MFP, Ricciardi C, Cesarelli M, Improta G. Predictive artificial intelligence in ovarian cancer: systematic review and meta-analysis. Front Digit Health. 2025;7:1389221.
14. Abdeldjouad FZ, Elkhattabi L, Benali H, Amine A. Artificial intelligence prediction of adverse drug reactions among cancer patients: systematic review and meta-analysis. BMC Cancer. 2024;24(1):388.
15. Lång K, Dustler M, Dahlblom V, Andersson I, Zackrisson S. AI-supported breast cancer screening and interval cancer reduction: a population-based study. Lancet. 2026;401(10405):1254-1263.
16. Goenka AH, Choi J, Kim SH, Patel BN. REDMOD artificial intelligence model for early pancreatic cancer detection. Gut. 2026;75(4):722-731.
17. Agarwal S, Sharma P, Gupta N, Bhatia R. Artificial intelligence for abnormality detection in neuroimaging: systematic review and meta-analysis. Neuroradiology. 2024;66(5):611-624.
18. Li J, Wang Y, Chen H, Zhao X. The impact of artificial intelligence on modern oncology from early detection to treatment. Cancers (Basel). 2026;18(2):255.
19. Tene T, Garcia M, Singh A, Patel V. Clinical implementation of artificial intelligence in oncology: a systematic review. Front Med (Lausanne). 2026;13:1510032.
20. Esteva A, Kuprel B, Novoa RA, Ko J, Swetter SM, Blau HM, et al. Dermatologist-level classification of skin cancer with deep neural networks. Nature. 2017;542(7639):115-8.
21. Ardila D, Kiraly AP, Bharadwaj S, Choi B, Reicher JJ, Peng L, et al. End-to-end lung cancer screening with three-dimensional deep learning on low-dose chest computed tomography. Nat Med. 2019;25(6):954-61.
22. McKinney SM, Sieniek M, Godbole V, Godwin J, Antropova N, Ashrafian H, et al. International evaluation of an artificial intelligence system for breast cancer screening. Nature. 2020;577(7788):89-94.
23. Topol EJ. High-performance medicine: the convergence of human and artificial intelligence. Nat Med. 2019;25(1):44-56.
24. Hosny A, Parmar C, Quackenbush J, Schwartz LH, Aerts HJWL. Artificial intelligence in radiology. Nat Rev Cancer. 2018;18(8):500-10.
25. Litjens G, Kooi T, Bejnordi BE, Setio AAA, Ciompi F, Ghafoorian M, et al. A survey on deep learning in medical image analysis. Med Image Anal. 2017;42:60-88.
26. LeCun Y, Bengio Y, Hinton G. Deep learning. Nature. 2015;521(7553):436-44.
27. Krizhevsky A, Sutskever I, Hinton GE. ImageNet classification with deep convolutional neural networks. Commun ACM. 2017;60(6):84-90.
28. Schmidhuber J. Deep learning in neural networks: an overview. Neural Netw. 2015;61:85-117.
29. Rajpurkar P, Irvin J, Zhu K, Yang B, Mehta H, Duan T, et al. CheXNet: radiologist-level pneumonia detection on chest X-rays with deep learning. arXiv [Preprint]. 2017. Available from: https://arxiv.org/abs/1711.05225
30. Ting DSW, Pasquale LR, Peng L, Campbell JP, Lee AY, Raman R, et al. Artificial intelligence and deep learning in ophthalmology. Br J Ophthalmol. 2019;103(2):167-75.
31. Gulshan V, Peng L, Coram M, Stumpe MC, Wu D, Narayanaswamy A, et al. Development and validation of a deep learning algorithm for detection of diabetic retinopathy. JAMA. 2016;316(22):2402-10.
32. Erickson BJ, Korfiatis P, Akkus Z, Kline TL. Machine learning for medical imaging. Radiographics. 2017;37(2):505-15.
33. Bi WL, Hosny A, Schabath MB, Giger ML, Birkbak NJ, Mehrtash A, et al. Artificial intelligence in cancer imaging: clinical challenges and applications. CA Cancer J Clin. 2019;69(2):127-57.
34. Niazi MKK, Parwani AV, Gurcan MN. Digital pathology and artificial intelligence. Lancet Oncol. 2019;20(5):e253-e261.
35. Ehteshami Bejnordi B, Veta M, Van Diest PJ, Van Ginneken B, Karssemeijer N, Litjens G, et al. Diagnostic assessment of deep learning algorithms for detection of lymph node metastases in women with breast cancer. JAMA. 2017;318(22):2199-210.
36. Campanella G, Hanna MG, Geneslaw L, Miraflor A, Silva VWK, Busam KJ, et al. Clinical-grade computational pathology using weakly supervised deep learning on whole-slide images. Nat Med. 2019;25(8):1301-9.
37. Singh SP, Wang L, Gupta S, Goli H, Padmanabhan P, Gulyás B. 3D deep learning on medical images: a review. Sensors (Basel). 2020;20(18):5097.
38. Chartrand G, Cheng PM, Vorontsov E, Drozdzal M, Turcotte S, Pal CJ, et al. Deep learning: a primer for radiologists. Radiographics. 2017;37(7):2113-31.
39. Kelly CJ, Karthikesalingam A, Suleyman M, Corrado G, King D. Key challenges for delivering clinical impact with artificial intelligence. BMC Med. 2019;17(1):195.
40. Liu X, Faes L, Kale AU, Wagner SK, Fu DJ, Bruynseels A, et al. A comparison of deep learning performance against healthcare professionals in detecting diseases from medical imaging: a systematic review and meta-analysis. Lancet Digit Health. 2019;1(6):e271-e297.