Open research questions in AI in cancer detection

What the literature leaves open

• The heterogeneity of tissue source between pre-treatment and posttreatment samples. The need for a computationally efficient workflow that can be implemented on a standard computer. The challenge of comparing QuTILs with other algorithms, such as CNN11, which is only available for outdated QuPath versions.

  QuTILs: open-source image-based infiltrating immune cell detection for research application · 2026 · DOI

• The need for a computationally efficient, open-source workflow for sTIL identification from digital H&E images for research application. The lack of a readily accessible resource for researchers to determine estimated TILs from digital pathology images.

  QuTILs: open-source image-based infiltrating immune cell detection for research application · 2026 · DOI

• This study proposed a segmentation-guided CNN and Vision Transformer feature fusion framework for multi-class breast ultrasound image classification. The main objective was to address the low rate of classification performance and to improve the classification of normal, benign, and malignant breast ultrasound images by first localizing the lesion region and then extracting more informative and complementary deep features from the focused region of interest. For lesion localization, FET UNet was used because of its ability to combine convolutional feature learning with transformer-based contextual modelling. The segmentation stage helped reduce the influence of irrelevant background tissue, speckle noise, and surrounding anatomical structures, thereby providing cleaner and more diagnostically meaningful inputs for the classi- fication stage. After segmentation, ResNet50 and ViT-based feature extractors were used to capture complementary infor- mation from the lesion-focused ultrasound images. The ResNet50 branch extracted local texture, boundary, and structural features, while the ViT branch captured global contextual relationships and long-range spatial dependencies. These features were fused to form a more robust feature representation and were then used to develop a deep neural network classifier. The proposed framework achieved strong classification performance, with an overall accuracy of 95.11%, macro sensitivity of 95.67%, macro specificity of 97.64%, and macro F1-score of 94.46%. The ROC and precision-recall analy- ses further confirmed the robustness of the proposed approach, with high class-wise AUC and average precision values across normal, benign, and malignant categories. Although the proposed framework demonstrated strong performance under cross-validation on the benchmark BUSI dataset, its generalizability to independently collected datasets remains unverified. Differences in ultrasound acqui- sition devices, imaging protocols, image quality, patient populations, class distributions, and annotation procedures may introduce domain shifts and affect its performance on external data. Therefore, future studies should evaluate the framework on independent breast ultrasound datasets containing comparable diagnostic categories and investigate the effects of domain-related differences on classification performance. If substantial domain-related performance deg- radation is observed, domain-adaptation and domain-generalization strategies, such as domain-adversarial training, maximum mean discrepancy-based feature alignment, correlation alignment, self-supervised pretraining, and test- time adaptation, could be incorporated into the framework to improve its robustness and generalizability across inde- pendently collected datasets. Table 2. Performance of different methods of the ablation study.

  A segmentation-guided CNN–Vision transformer feature fusion framework for multi-class breast ultrasound image classification · 2026 · DOI

• The challenge of automatically diagnosing breast cancer through advanced techniques and real images. The difficulty in comparing the performance of different classifiers and regression models for breast cancer diagnosis. The need to address the issue of late detection of breast cancer and its associated effects.

  Advancing Breast Cancer Diagnosis through Breast Mass Images, Machine Learning, and Regression Models · 2024 · DOI

• more algorithms and combinations of features can be considered for the precise, rapid, and effective classification and diagnosis of breast cancer images - early automatic detection through screening and prompt evaluation of any abnormalities are key factors in improving the outcomes for individuals

  Advancing Breast Cancer Diagnosis through Breast Mass Images, Machine Learning, and Regression Models · 2024 · DOI

• The low accuracy observed in this setting confirms that global color features are insufficient for fine-grained cancer subtype differentiation, as these distinctions rely primarily on architectural and morphological patterns rather than chromatic differences.

  Beyond Morphology: Quantifying the Diagnostic Power of Color Features in Cancer Classification · 2026

• comparison with other recently published methods, - further validation of the proposed framework

  A segmentation-guided CNN–Vision transformer feature fusion framework for multi-class breast ultrasound image classification · 2026 · DOI

• The main limitation of this study is the dataset's class imbalance, and while weighted training helped mitigate this, further research and additional data are needed to guarantee model generalizability.

  Potential of Artificial Intelligence Algorithms for Identification of Relevant Diagnostic and Prognostic Biomarkers of Early-Stage Liver Cancer · 2026

• The study essentially focuses on addressing a key limitation of deep learning in the context of its limited ability to generalize effectively across diverse breast cancer datasets and image formats.

  Breast cancer image classification using deep learning augmented with attention mechanism · 2026 · DOI

• Usually, tumors will vary widely in their characteristics, behavior, initial position of tumors and prognosis, requiring accurate and timely identification to initiate optimal patient treatment.

  Deep learning based identification and analysis of brain and breast tumours using VGG16 ResNet50 and EfficientNetB0 · 2026 · DOI

• However, its diagnostic concordance in thyroid cytopathology, particularly under real-world conditions, remains insufficiently validated.

  Diagnostic concordance of remote WSI-based thyroid cytology compared to conventional cytological diagnosis: a pilot study · 2026 · DOI

• Notably, few studies have addressed real-world deployment challenges, and recent advances, such as explainable AI (XAI), federated learning, and multimodal approaches, remain underrepresented in this context.

  Leveraging AI and data science across the cervical cancer care continuum in developing economies · 2025 · DOI

• Overall, AI-based mammographic risk models generally improve discrimination versus classical models and are being externally validated; however, evidence remains heterogeneous across molecular subtypes, with signals strongest for ER-positive disease and limited data for fast-growing and interval cancers.

  Artificial Intelligence-Driven Personalization in Breast Cancer Screening: From Population Models to Individualized Protocols · 2025 · DOI

• Evaluating the pipeline on a larger number of whole-slide images. Applying the pipeline to other types of cancer or datasets. Improving the accuracy of the pipeline using additional training data or techniques.

  An integrated automated deep learning framework for annotating tumor-infiltrating lymphocytes in lung adenocarcinoma pathology · 2026 · DOI

• Acquiring large-scale, fully annotated datasets remains a major obstacle. Supervised learning using extensively annotated whole-slide images remains the primary method for developing high-precision computational models.

  An integrated automated deep learning framework for annotating tumor-infiltrating lymphocytes in lung adenocarcinoma pathology · 2026 · DOI

• The study identifies a gap between theoretical and practical deployment of AI systems. The study notes that dataset diversity, ethical issues, and real-world clinical validation are still problems.

  Artificial Intelligence in Medical Imaging: A Critical Review of Methods, Applications, and Clinical Implementation · 2026 · DOI

• Breast ultrasound examinations can be time-consuming and require high levels of expertise. Diagnostic accuracy can be limited in women with dense breast tissue.

  Artificial intelligence–assisted breast ultrasound: modest AUROC improvement and shorter interpretation time without significant change in diagnostic accuracy · 2026 · DOI

• There is a need to improve diagnostic discrimination and interpretation efficiency in breast ultrasound examinations. There is a lack of studies evaluating the effectiveness of AI assistance in breast ultrasound examinations.

  Artificial intelligence–assisted breast ultrasound: modest AUROC improvement and shorter interpretation time without significant change in diagnostic accuracy · 2026 · DOI

• There is a need for accurate and reliable decision-support systems for breast cancer diagnosis - The current methods of histopathological analysis are time-consuming and rely heavily on the skill and experience of the pathologist

  Transformers meet CNNs for insights into breast mass classification from histopathological images · 2026 · DOI

• • Investigated three deep learning approaches • Deep learning methods • Performance depends on the availability of for breast ultrasound (BUS) lesion detection: demonstrated overall superior labeled training data, including negative samples. patch-based LeNet, U-Net, and transfer detection performance across both • Evaluation is limited to 2D ultrasound lesion learning with pretrained FCN-AlexNet. datasets in terms of True Positive detection only, without downstream • Performance compared with four state-of- Fraction (TPF), false positives per segmentation or classification. the-art conventional methods: Radial Gradient image, and F-measure. • Dataset sizes are relatively small, and multi-center Index, Multifractal Filtering, Rule-based • FCN-AlexNet achieved the best validation is not explored. Yap et al. (2017) Region Ranking, and Deformable Part results on Dataset A, while patch- • The lack of a unified public benchmark remains a Models (DPM). based LeNet performed best on challenge. • Evaluated on two ultrasound datasets acquired Dataset B. using different imaging systems. • The study highlighted the adaptability and robustness of learning-based lesion detection methods across heterogeneous datasets. • Proposed a supervised breast ultrasound • Achieved competitive • The method is not fully automated, requiring (BUS) tumor segmentation method based on segmentation performance on a manual ROI selection by the operator. semantic classification and superpixel merging. BUS dataset of 320 cases, • Performance strongly depends on classifier • ROI is manually selected, followed by image producing tumor contours close to accuracy and dataset-specific training, limiting enhancement (histogram equalization, bilateral hand-labeled annotations. generalization across imaging devices. filtering, pyramid mean shift). • Reported an average F1-score of • While effective, results are inferior to fully Huang et al. (2020) • SLIC superpixels are generated, semantic 89.87% ± 4.05% and favorable TP/ supervised deep learning models (e.g., FCN) features extracted, and a Bag-of-Words (BoW) FP rates compared with five trained with pixel-level annotations. model constructed. existing approaches. • Scalability to large datasets and real-time clinical • Initial classification is performed using a • Demonstrated that semantic deployment was not evaluated. Backpropagation Neural Network (BPNN) and superpixel classification can refined using K-Nearest Neighbors (KNN) effectively guide tumor boundary reclassification. delineation. • Proposed a novel semantic segmentation • Demonstrated superior • Performance remains dependent on the quality network incorporating a channel attention segmentation performance and diversity of annotated training data. module with Multiscale Grid Average Pooling compared to existing deep learning • Like other deep learning approaches, the model (MSGRAP) to improve breast cancer approaches on a public requires pixel-level ground truth, increasing segmentation in ultrasound images. BUS dataset. annotation cost. Lee et al. (2020) • The attention module captures both global and • The proposed channel attention • Generalization to ultrasound images from local spatial information, overcoming module significantly enhanced different acquisition devices or clinical settings limitations of standard convolution. Network segmentation accuracy, showing was not extensively analyzed. architecture was optimized through ablation better semantic consistency and studies and evaluated against FCN, SegNet, boundary delineation of breast U-Net, and other state-of-the-art deep cancer regions. learning models. • Hybrid framework integrating Dense • Achieved high classification • Model interpretability remains limited due to the Convolutional Network (DenseNet) for accuracy of 99.32% on BreakHis black-box nature of deep learning. hierarchical feature extraction with Flower and 96% on BACH datasets, • Validation is restricted to benchmark datasets; Pollination Algorithm (FPA) for optimized outperforming existing state-ofreal-world clinical deployment and large-scale Wakili et al. (2025) feature selection on histopathological images the-art methods. multi-center validation are not explored. (BreakHis and BACH datasets). • Demonstrated improved • Computational cost of hybrid optimization may generalization, reduced increase training time.

  Transformers meet CNNs for insights into breast mass classification from histopathological images · 2026 · DOI

• The model's performance was limited by overfitting and low precision in some experiments. The dataset used was limited, which may affect the generalizability of the results. The model's ability to detect tumors in images with noise or artifacts is not evaluated.

  Breast cancer detection using deep learning techniques: An investigation using the CBIS-DDSM dataset and customized neural network model, ResNetV2 and YOLO · 2026 · DOI

• To develop more accurate and reliable XAI methods for mammographic screening. To improve the integration of AI into diverse clinical workflows. To address the data-related issues and develop large, diverse, and high-quality datasets with accurate labeling.

  From Black Box to Glass Box: A Survey of Explainable AI in Mammographic Screening · 2026 · DOI

• The gap between technical innovation and clinical utility in mammographic screening. The need for transparent AI in medicine to improve breast cancer diagnosis. The lack of large, diverse, and high-quality datasets with accurate labeling for developing reliable AI models.

  From Black Box to Glass Box: A Survey of Explainable AI in Mammographic Screening · 2026 · DOI

• High false positive and false negative rates of traditional screening methods. Limited access to large and diverse datasets. Domain adaptation issues due to equipment differences and/or differences in the patient population.

  Explainable Deep Learning Framework for Breast Cancer Classification · 2026 · DOI

• Further development of deep learning models for breast cancer classification. Investigation of the use of Explainable Artificial Intelligence (XAI) in medical image analysis. Exploration of the potential applications of the proposed framework in clinical practice.

  Explainable Deep Learning Framework for Breast Cancer Classification · 2026 · DOI

Most-cited papers in AI in cancer detection

• Diagnostic Assessment of Deep Learning Algorithms for Detection of Lymph Node Metastases in Women With Breast Cancer · JAMA · 2017 · 2,894 citations
• Segment anything in medical images · Nature Communications · 2024 · 2,440 citations
• Towards a general-purpose foundation model for computational pathology · Nature Medicine · 2024 · 1,185 citations
• Medical image segmentation using deep learning: A survey · IET Image Processing · 2022 · 791 citations
• A whole-slide foundation model for digital pathology from real-world data · Nature · 2024 · 656 citations
• Breast Cancer Diagnosis and Prognosis Via Linear Programming · Operations Research · 1995 · 586 citations
• ASF-YOLO: A novel YOLO model with attentional scale sequence fusion for cell instance segmentation · Image and Vision Computing · 2024 · 446 citations
• A pathology foundation model for cancer diagnosis and prognosis prediction · Nature · 2024 · 443 citations
• A foundation model for clinical-grade computational pathology and rare cancers detection · Nature Medicine · 2024 · 400 citations
• Multimodal data fusion for cancer biomarker discovery with deep learning · Nature Machine Intelligence · 2023 · 387 citations

Most recent work

• Interval cancer, sensitivity, and specificity comparing AI-supported mammography screening with standard double reading without AI in the MASAI study: a randomised, controlled, non-inferiority, single-blinded, population-based, screening-accuracy trial · The Lancet · 2026
• An agentic framework for autonomous scientific discovery in cancer pathology · Nature Medicine · 2026
• Geometric multi-instance learning for weakly supervised gastric cancer segmentation · npj Digital Medicine · 2026
• AI-based triage and decision support in mammography and digital tomosynthesis for breast cancer screening: a paired, noninferiority trial · Nature Medicine · 2026
• Foundation model embeddings for multimodal oncology data integration · npj Digital Medicine · 2026
• Structure-aware generalization for heterogeneous histopathology via prototype-based multiple instance learning · npj Digital Medicine · 2026
• Commercially Available Artificial Intelligence Score on Preoperative Mammography for Prediction of Future Breast Cancer After DCIS Treatment · American Journal of Roentgenology · 2026
• HED-Net: a hybrid ensemble deep learning framework for breast ultrasound image classification · Frontiers in Artificial Intelligence · 2026
• UroFusion-X: a unified multimodal deep learning framework for robust diagnosis, subtyping, and prognosis of urological cancers · npj Digital Medicine · 2026
• Mammography-based artificial intelligence model for predicting axillary lymph node status after neoadjuvant therapy in breast cancer · European Radiology · 2026

Related topics in Computer Science