Breast cancer is the most-common invasive cancer among women. It affects 14% of women worldwide and every year, over half a million women die of it. That makes breast cancer one of the largest medical problems faced today.
Taking treatment decisions and curing this disease requires, at first place, diagnosing it.
However, due to time constraints and the existence of very small metastases on individual slides, human pathologists sometimes fail to detect cancerous cells: they miss about 20% of breast cancers in mammograms, even in countries where screening mammograms are reviewed by two radiologists.
In the last years, the introduction of deep learning and convolutional neural networks (CNNs) in medical image analysis has prevaded the field of automated breast cancer detection in digital mammography.
Researchers have created an AI diagnostic tool that helps doctors detect breast cancer.
The man-made neural network is trained with more than a million mammography images, and the program is getting “smarter” as it reviews more and more data.
The AI tool was taught to analyze even the small patches, spot changes that are invisible to the human, and then make a map of the areas that are most at risk.
In addition to its success in the detection of abnormality as well as an average radiologist, this AI invention has won two main challenges concerning the efficiency and consistence of the results: Reducing false negatives by 9% and false positives by 5.7%.
At this level, you might be asking yourself whether or not this program will replace doctors when it comes to detecting breast cancer. The answer is definitely a NO!
In fact, the knowledge we’re getting when the program points out cancer to us, is the knowledge of many pathologists who contributed to training the machine on a model. It’s mainly human knowledge.
All we said above, highlights the fact that we must combine the strengths of human radiologists with the results of these performing systems in order to provide the most accurate diagnosis for the patient and thus allow clinicians to have significantly more time to deal with curing him.
“AI detected pixel-level changes in tissue invisible to the human eye, while humans used forms of reasoning not available to AI,” says senior study author Krzysztof J. Geras, PhD, assistant professor in the Department of Radiology at NYU Langone and an affiliated faculty member at the NYU Center for Data Science. “The ultimate goal of our work is to augment, not replace, human radiologists.”
Share your thoughts
Episode 2: Buying a Soccer Team
As we are progressing into a world where sports have become a vital part of our lives, it has also become a hot market for investors to gain greater income. We can also see that there has been a surge in sports viewership which leads to more tournaments. Capitalizing on them has become a difficult task for an investor. So a programmers team have come up with a strategy to build the best team whilst keeping in mind the investor’s budget which has a limit of 1 billion Euros relying on Machine learning algorithms to detect potential recruits for their clubs and potential assets for the investor to optimize their market gains.
How does that work ?
The team have a FIFA dataset in which there are few columns named — rating, release clause, and wages. They built 2 models using Supervised Learning on Rating variables and made it a classification problem by splitting this variable into 2 classes: greater than or equal to 70 and less than 70 ratings. They chose 70 as their threshold as most of the major clubs have only players with ratings greater than 70. In the second model, the rating class was obtained from previous best classifiers instead of « actual rating » and the release-clause was combined with annual wages to determine the cost to investors.
The dataset used is FIFA 19 and FIFA 20 data which contain 18k+ unique player ratings with 100+ different attributes for each player:
- Player positions, with the role in the club and in the national team.
- Player attributes with statistics as Attacking, Skills, Defense, Mentality, GK Skills, etc.
- Player personal data like Nationality, Club, DateOfBirth, Wage, Salary, etc.
Example: Bangladesh National Team
Due to the miscoordination among the players recently, the national team of Bangladesh slipped to their worst ever FIFA ranking of 194th. The ultimate success of a team depends upon its player selection. Generally, the coach and the management team select the best 11 players for each match from a pool of players by evaluating various attributes of the players. Now, using these Machine Learning algorithms, a better player classification technique is proposed and the soccer team is formed automatically of the best player combination possible. The results achieved based on the performance attributes would help the coach and management team to build up a more successful soccer team.
The constructed 2 models that utilize Machine learning algorithms benefit investors while simultaneously select the rightfully classified players as good performers and then regress them in the budget of the investor. Ultimately, the narrowed down selection process of a player within a club is rather better than selecting at random.