Machine learning model for stratification of early diabetic foot complication using thermogram images

A system, comprising:

• a processor;
• and
• a memory that includes instructions, that when executed by the processor perform operations including:
• receiving a thermogram of a foot;
• identifying, via a machine learning model, a risk factor of diabetic foot complications including diabetic ulceration on the foot and at least one of heart disease, stroke, renal failure and blindness, wherein the machine learning model includes poolwise-dense connections in which tanh-activated feature maps from multiple Self Inverted Bottleneck Residual (IBR) blocks are concatenated via adaptive average pooling to feed a Self Multilayer Perceptron (SelfMLP) classification layer, thereby enabling feature reuse, strengthened propagation, and normalized processing across layers to improve accuracy of severity stratification into mild, moderate, or severe categories;
• outputting, from the machine learning model, the risk factor as a diagnosis that stratifies the risk into one of the mild, moderate, or severe categories;
• and
• applying a prophylactic intervention selected from off-loading therapy, pharmacotherapy, or lifestyle modification to prevent progression to diabetic foot ulceration when the risk is stratified as moderate or severe.

A device including computer-readable instructions, that when executed by a processor, perform operations comprising:

• receiving a thermogram of a foot;
• identifying, via a machine learning model, a risk factor of diabetic foot complications including diabetic ulceration on the foot and at least one of heart disease, stroke, renal failure and blindness, wherein the machine learning model includes poolwise-dense connections in which tanh-activated feature maps from multiple Self Inverted Bottleneck Residual (IBR) blocks are concatenated via adaptive average pooling to feed a Self Multilayer Perceptron (SelfMLP) classification layer, thereby enabling feature reuse, strengthened propagation, and normalized processing across layers to improve accuracy of severity stratification into mild, moderate, or severe categories;
• outputting, from the machine learning model, the risk factor as a diagnosis that stratifies the risk into one of the mild, moderate, or severe categories;
• and
• applying a prophylactic intervention selected from off-loading therapy, pharmacotherapy, or lifestyle modification to prevent progression to diabetic foot ulceration when the risk is stratified as moderate or severe.