Diagnosis of tuberculosis (TB) is time-consuming, cumbersome, and expensive. Moreover, there is a lack of real-time monitoring of screening and testing as well as data management and storage. Serological screening point-of-care tests, which are rapid and affordable, have been viewed as a desirable method for TB diagnosis for a long time, although they cannot be used to confirm the disease. Three novel antigens of mycobacterium TB (MTB), the causative agent of TB, have been considered for the colorimetric diagnosis. The immunochromatic flowthrough test (ICT) devices were developed to screen the suspected cases of active TB with high sensitivity and specificity. In this work, using these ICT devices, we have now developed an image sensing method based on dataset of images and trained a model to create a custom-made phone application for the accurate detection of TB with real-time reporting. The image sensing of the colorimetric outcome was integrated with different classifications, of which feedforward neural network (FNN) allowed us to make predictions with an overall accuracy of similar to 82%, and this is on par with the results of existing literature. With a sensitivity of 87%, a specificity of 82%, an AUC score of 0.84, and an F1-score of 81% (No TB) and 82% (with TB), the suggested approach demonstrates enhanced efficiency compared to naked eye results. This image sensing technique can significantly reduce the possibility of errors resulting from visual results and color ambiguity.