How can we improve breast cancer prediction thanks to machine learning?
Breast cancer prediction is a diagnosis tool. Oncologists and medical staff face the challenge of identifying breast cancer as soon as possible. An early diagnosis increases survival chances. TADA can help them diagnose malignant tumors quicker.
Industry
Healthcare
Project Duration and Effort
One week
Type of Prediction
Binary Classification
Customer Benefits
96% accuracy in identifying cancer-causing cell nuclei with TADA versus 79% by clinicians.
No need to be an experienced physician, substantial accuracy available for senior and junior physicians alike.
Speed, once the tool is in place, TADA’s analysis takes a few minutes.
TADA improves early cancer detection by 17%.
Problems to solve
Breast cancer is one of the most common cancers today in women. A breast mass in patients means a tumor. It does not necessarily imply a malignant one. Hence, American oncologists perform a fine needle aspirate (FNA) on the cancer patient.
Then, they examine the resulting cells and extract the cells nuclei features. The most critical step is this feature extraction.
The goal is to select elements of this image that one can measure for further computational analysis. And at the same time, the measures should be representative of cancer. We aim to use elements of the image measured as either a diagnostic or a prognostic indicator. The early detection of cancer is an efficient means to improve the patient’s survival rate.
The typical features extracted are:
radius
texture
perimeter
area
smoothness
compactness
concavity (severity of concave portions of the contour)
concave points (number of concave portions of the contour)
symmetry
fractal dimension
and several figures are computed for each feature.
The diagnosis of cancer has been mostly dependent on traditional approaches, using trained professionals’ expertise. However, a senior trained professional is not always available.
TADA’s Machine Learning approach can help automate, in part, the cancer risk prediction. Thus senior and junior professionals alike get access to the same analyzed data from cancer patients. They can provide a better, quicker diagnosis, hence improving survival rates.
Objectives
Make the distinction between benign and malignant tumors after an FNA rapidly.
Support, improve, and reassure oncologists in their diagnoses.
Improve the accuracy of breast cancer prediction.
It poses the following oncology question:
Can cancer prediction distinguish malignant from benign tumors?
Solution
A Fine Needle Aspiration Biopsy (FNA) is a biopsy that produces a fast, reliable, and economic evaluation of tumor lesions. It is a minimally invasive scheme that utilizes a fine needle to aspirate tissue from mass lesions. It expedites the sequence between the diagnosis and the beginning of therapy for breast cancer.
FNA is ideally conducted by an expert medical biologist who can follow with prompt microscopic examination. Research indicates that the most experienced physicians can diagnose breast cancer using FNA with a 79% accuracy.
TADA has selected the following four main criteria out of the thirty available in the dataset. They approximately bear the same weight in the decision to identify breast cancer:
1. the worst radius
2. the worst number of concave points around the contour
3. the radius standard deviation
4. the smoothness
A 17% improvement in breast cancer predictions happens through TADA (from 79% to 96%). The artificial intelligence tool distinguishes benign from malignant tumors by providing a probability of the tumor being one or the other. It can also help the oncologist understand how each element measured impacts the diagnosis. For instance, it can prove the relationship between the tumor’s overall dimension and breast cancer chances.
Customer Benefits
In one week, oncologists gained significant support in their cancer diagnosis and their fight against breast cancer by:
- Making the difference between benign and malignant cancer quickly.
- Diagnosing malignant cancers with a 96% accuracy.
- Obtain an immediate “what-if” analysis linking the tumor’s characteristics and cancer.
Talk to us about how you can make sense of your data and achieve success.
