EfficientNet Models for Detection of Anemia Disorder using Palm Images

Anemia is a disorder that causes the body to have insufficient red blood cells or hemoglobin resulting in a decrease in the blood's ability to carry oxygen to the body's tissues. Elevated pulse, weariness, skin pallor, shortness of breath, lightheadedness, and dizziness are a few symptoms that may appear on the onset of Anemia. The underlying diagnosis affects the course of treatment. To treat an iron deficit patient, iron supplements are utilized. Low vitamin levels may be treated with vitamin B supplements. This study explains the use of the Convolutional Neural Networks (CNN) to categorize hemoglobin levels using patient palm photographs. Anemia and non-Anemia are classified according to the hemoglobin levels. ML-assisted illness diagnosis is more accurate, less time-consuming, and requires less additional labour than traditional methods of diagnosis. EfficientNetB0 model with softmax activation function and Adam optimizer achieved the best results, with an accuracy of 97.52% and an F1 score of 97%. The full family of EfficientNet models, from B0 to B7 performed well with both sigmoid and softmax activation functions and Adam optimizer, with results consistently exceeding 92%. In contrast, the same models with linear activation function underperformed.