Algorithms for Tumor Detection
Detecting tumors in their early stages is critical to treating them as effectively as possible. However, the detection of tumors by traditional methods may be time-consuming and is sometimes inaccurate (1). Computer intelligence-based techniques may support physicians in the identification and classification of tumors. Several machine learning and deep learning algorithms for tumor detection and diagnosis have been established to this end.
The process of leveraging algorithms for tumor detection usually involves a number of steps. First, an image of the tumor area or suspected area is acquired. This can take the form of an X-ray, computed tomography scan, magnetic resonance imaging scan, or ultrasound. Next, the image is preprocessed to remove noise and improve its quality, after which the tumor region is segmented or extracted out from the rest of the image, either manually or algorithmically. Features are then extracted from the image, such as tumor size, texture, or shape, after which these are leveraged to classify the tumor as benign or malignant. This classification step can be achieved using various machine learning algorithms, including neural networks, decision trees, or support vector machines. Finally, a medical professional makes the final diagnosis of the type and severity of the tumor as informed by the automatic classification results.
Brain and other central nervous system tumors are among the most fatal cancers (2), making them particularly important to detect early and accurately. One research team has highlighted the importance in particular of leverage magnetic resonance imaging scans for the automatic diagnosis of brain tumors. Focusing on three different types of tumors, including a glioma, a meningioma, and a pituitary tumor, the team assessed three algorithms – k-nearest neighbor, support vector machine, and general regression neural network – for tumor detection and diagnosis (1). All methods demonstrably reduced the diagnostic time and increased the accuracy of tumor diagnosis by leveraging a publicly available dataset, performing image pre-processing, extracting out image features, and carrying out principal component analysis methods. In the end, the accuracies of the algorithms used in tumor diagnosis for the KNN, SVM, and GRNN were 97%, 96.24%, and 94.7%, respectively. The KNN was finally considered the best algorithm.
Rather than focusing on different algorithms, another study sought to probe differences across different deep learning techniques (3). The team compared the pre-trained InceptionResNetv2, Inceptionv3, and ResNet50 machine learning models, as well as their newly introduced BRAIN-TUMOR-net, on three different publicly available magnetic resonance imaging datasets. Simulation results revealed that the BRAIN-TUMOR-net achieved the highest accuracy, thus representing a very viable clinical diagnostic option.
Finally, a 2023 study focused on brain tumors proposed several machine learning approaches and two different deep learning methods for distinguishing between three types of tumor, i.e., a glioma, a meningioma, and a pituitary gland tumor, as well as healthy tumor-less brains. These algorithms used magnetic resonance brain images to help physicians with early-stage tumor detection (4). Results demonstrated that a proposed two-dimensional convolutional neural network achieved exemplary performance and optimal execution time without latency. The proposed network was not as complex as the auto-encoder network they used in comparison, and can readily be harnessed by radiologists and other medical professionals for the detection of brain tumors.
Advancements in algorithms and other technological methods have buoyed tumor detection and diagnoses by health professionals in many ways. Additional research and targeted algorithms will no doubt continue to improve our ability to catch tumors and begin treatment in earlier stages.
References
1. Refaat FM, Gouda MM, Omar M. Detection and Classification of Brain Tumor Using Machine Learning Algorithms. Biomed Pharmacol J. 2022 Dec 20;15(4):2381–97. doi: 10.26452/ijrps.v10i3.1442
2. Miller KD, Ostrom QT, Kruchko C, Patil N, Tihan T, Cioffi G, et al. Brain and other central nervous system tumor statistics, 2021. CA Cancer J Clin. 2021;
3. Taher F, Shoaib MR, Emara HM, Abdelwahab KM, Abd El-Samie FE, Haweel MT. Efficient framework for brain tumor detection using different deep learning techniques. Front Public Heal. 2022 Dec 1;10:3374. doi: 10.3389/fpubh.2022.959667
4. Saeedi S, Rezayi S, Keshavarz H, R. Niakan Kalhori S. MRI-based brain tumor detection using convolutional deep learning methods and chosen machine learning techniques. BMC Med Inform Decis Mak. 2023 Jan 23;23(1):1–17. doi: 10.1186/s12911-023-02114-6