Anatomical landmark detection and identification from digital radiography images containing severe skeletal deformations

A system for detecting anatomical landmarks in abnormal subjects, the system comprising:

• a memory storing instructions;
• one or more communication interfaces;
• and
• one or more hardware processors coupled to the memory via the one or more communication interfaces, wherein the one or more hardware processors are configured by the instructions to:
• obtain one or more digital radiography (DR) images of an abnormal subject, wherein the one or more DR images comprise one or more deformed structures of the abnormal subject, and wherein the one or more deformed structures are representative of one or more abnormalities observed in the abnormal subject;
• filter, using a first trained neural network executed by the one or more hardware processors, one or more artifacts from the obtained one or more DR images to obtain a pre-processed DR image comprising the one or more deformed structures of the abnormal subject;
• generate, using a second trained neural network executed by the one or more hardware processors, a heat map based on the pre-processed DR image, wherein the heat map comprises one or more locations of corresponding one or more probable anatomical landmarks in the pre-processed DR image;
• generate, by a trained classifier executed by the one or more hardware processors, a first probability score for each of the one or more locations of the corresponding one or more probable anatomical landmarks in the pre-processed DR image;
• perform a first comparison of the first probability score for each of the one or more locations of the corresponding one or more probable anatomical landmarks in the pre-processed DR image with a pre-defined threshold;
• and
• select, based on the first comparison, a subset of the corresponding one or more probable anatomical landmarks in the pre-processed DR image serving as a set of detected anatomical landmarks.

A computer program product comprising a non-transitory computer readable medium having a computer readable program embodied therein, wherein the computer readable program, when executed by one or more hardware processors, causes the one or more hardware processors to detect anatomical landmarks in abnormal subjects by:

• obtaining, via one or more hardware processors, one or more digital radiography (DR) images of an abnormal subject, wherein the one or more DR images comprise one or more deformed structures of the abnormal subject, and wherein the one or more deformed structures are representative of one or more abnormalities observed in the abnormal subject;
• filtering, using a first trained neural network executed by the one or more hardware processors, one or more artifacts from the obtained one or more DR images to obtain a pre-processed DR image comprising the one or more deformed structures of the abnormal subject;
• generating, using a second trained neural network executed by the one or more hardware processors, a heat map based on the pre-processed DR image, wherein the heat map comprises one or more locations of corresponding one or more probable anatomical landmarks in the pre-processed DR image;
• generating, by a trained classifier executed by the one or more hardware processors, a first probability score for each of the one or more locations of the corresponding one or more probable anatomical landmarks;
• performing, via the one or more hardware processors, a first comparison of the first probability score of each of the one or more locations with a pre-defined threshold;
• and
• selecting, via the one or more hardware processors, based on the first comparison, a subset of the corresponding one or more probable anatomical landmarks serving as a set of detected anatomical landmarks.