Diagnose Cancer, Just in 2 hours by an Algorithm

To detect the cancer in two hour is a great news. We all should rejoice at this glad tidings. All over the world, doctors are relying on an algorithm to do progress in the perfect diagnosis of cancer. Algorithm helps doctor to accelerate the diagnosis. It helps them in planning treatments. The goal of using this technique is to check more patients in more time with precision. What it takes to become a doctor one can understand conceptually that years of attending lectures in medical school, pile of books, reading journals, countless job hours are the part of this field. But the way, the medicine is learnt by Algorithm is less instinctive.

In order to understand more clearly that how these patterns are learnt by algorithm and what are the drawback that still sneaks within this technology, quartz partnered with the co-founder of medical Algorithm startup MD Leon Chen and Luke Oakden Rayner, to prepare two Algorithm and he tries to understand that how it will match with the medical professional as it learns. The presence of tumour module is detected by the one and the other determine the potential of malignancy in it.

AI detect cancer in two hours
Quartz

The artificial intelligence developed for medical use is a complicated pattern matching. The images of organs with tumours and tumour free are shown in an Algorithm. The task is to learn the patterns so that the two categories are differentiated.

Algorithm can almost show two hundred thousand images of benign, malignant and tumour free CT scans. These images are available in both 2D and 3D. If we want to check that how accurate the nodule detection Algorithm is to see the finding of tumour by a specialist by using recall should show the same result as it shows. The percentage of nodule and given number of false alarm set is explained by Recalls method.

AI cancer diagnosis
The nodule surrounded in red is extremely dense (bright), compared to the nodule surrounded in white, which is greyer. The previous could be a benign calcified nodule; the latter is sort of definitely cancerous. Image: Quartz

One could set the percentage of false alarm threshold higher or lower theoretically. This would also effect the impact the percentage of nodule caught. If we allow four false alarms per nodule, the percentage will go up. But one false alarms also indicates that it is an unnecessary test. Every doctor is comfortable with different level of algorithm.

LEAVE A REPLY

Please enter your comment!
Please enter your name here