Update README.md

We tackle the limitations of current Acute Myeloid Leukemia (AML) diagnostics by transferring knowledge from Acute Lymphoblastic Leukemia (ALL) classification models, thus addressing the critical need for improved and more accessible diagnostic tools for Acute Myeloid Leukemia detection. AML has poorer prognosis than ALL, with a 5-year relative survival rate of only 17–19% compared to ALL survival rates of up to 75%, making early and accurate detection of AML paramount.

The data used in this research comes from two sources, ALL images are from the Acute Lymphoblastic Leukemia (ALL) dataset, titled, C-NMC-2019(referred to as the "ALL Dataset"). AML images are taken from the Munich AML Morphology Dataset, titled, AML-Cytomorphology-LMU (referred to as the "AML Datset"). Both datasets were downloaded from The Cancer Imaging Archive (TCIA).

From the the ALL Dataset, we use 10,661 pre-classified, single-cell images. 7,272 images are classified as leukemic lymphoblasts (ALL), while the remaining images depict non-cancerous cells. These 7,272 ALL images are split into training, validation, and testing (held-out) sets in a 74:22:4 ratio, respectively. The ALL Dataset images are in bitmap format with a resolution of 450x450 pixels and are pre-segmented, meaning non-cell areas are black. The AML Dataset consists of expert-labeled single-cell images from blood smears of 100 patients diagnosed with AML at Munich University Hospital between 2014 and 2017. It also includes images from 100 patients without any indications of blood-cell malignancy (control group). These images are in TIFF format with a resolution of 400x400 pixels. Analysis of class balance reveals that 38% of the total images (10,989) show cancer, while 62% (18,117) are not.

README.md

@@ -1,3 +1,16 @@
----
-license: apache-2.0
----
+---
+license: apache-2.0
+language:
+- en
+metrics:
+- f1
+- precision
+- recall
+tags:
+- medical
+- cancer
+- ALL
+- AML
+- Transfer
+- Learning
+---