Automated pattern recognition and scoring method of histological images

1. A method for determining and scoring hepatocyte lobular inflammation in a liver biopsy from a subject, the method comprising:

providing a liver biopsy slide from said subject;

using a deep learning model:

determining a plurality of field of views (FOVs) on a digitized form off the liver biopsy slide; and

selecting a subset of the plurality of FOVs and identifying inflammatory cell patterns within the subset of the plurality of FOVs;

determining a hepatocyte lobular inflammation (LI) score for the liver biopsy slide based on the identified inflammatory cell patterns;

wherein the LI score ranges from 0 to 3 and LI=0 corresponds to no inflammation, LI=1 corresponds to mild inflammation, LI=2 corresponds to moderate inflammation, and LI=3 corresponds to severe inflammation.

2. The method according to claim 1 , wherein the deep learning model comprises a model trained based on training data obtained from liver biopsy imaging data comprising inflammatory cells and normal hepatocytes.

3. A method for diagnosing NASH in a subject, the method comprising:

providing a liver biopsy slide from said subject;

using a deep learning model:

determining a plurality of field of views (FOVs) on a digitized form of the liver biopsy slide; and

selecting a subset of the plurality of FOVs and identifying inflammatory cell patterns within the subset of the plurality of FOVs;

identifying patterns of hepatocyte ballooning (HB) and score hepatocyte ballooning;

determining a hepatocyte lobular inflammation (LI) score for the liver biopsy slide based on the identified inflammatory cell patterns;

wherein the LI score ranges from 0 to 3 and LI=0 corresponds to no inflammation, LI=1 corresponds to mild inflammation, LI=2 corresponds to moderate inflammation, and LI=3 corresponds to severe inflammation; and

wherein HB score ranges from 0 to 2 and HB=0 corresponds to no hepatocyte ballooning, HB=1 corresponds to moderate hepatocyte ballooning, and HB=2 corresponds to severe hepatocyte ballooning and wherein said HB score and said LI score are added to determine an Activity Index (AI), ranking from 0 to 5, wherein the subject is diagnosed as a NASH subject if AI≥2, with HB≥1 and LI≥1.

4. The method according to claim 3 wherein NASH is diagnosed when, the HB score is of at least 1, the LI score is of at least 1, and a steatosis score from the same subject is of at least 1.

5. A method for determining and scoring NASH fibrosis or liver fibrosis in a subject, the method comprising:

providing a liver biopsy slide from said subject;

using a deep learning model:

determining a plurality of field of views (FOVs) on a digitized form of the liver biopsy slide;

selecting a subset of the plurality of FOVs and identifying patterns of liver fibrosis and/or measuring collagen proportion area (CPA) within the subset of the plurality of FOVs;

determining a liver fibrosis score for the liver biopsy slide based on the identified patterns and/or measured proportions;

wherein the liver fibrosis (F) is scored as follows: F=0 corresponds to no liver fibrosis; F=1 corresponds to minimal liver fibrosis; F=2 corresponds to significant liver fibrosis; F=3 corresponds to moderate liver fibrosis; and F=4 corresponds to severe liver fibrosis.

6. The method according to claim 5 , wherein the deep learning model comprises a model trained based on training data obtained from liver biopsy imaging data comprising normal and collagen-containing tissues.

7. The method according to claim 6 , wherein the training data comprise data obtained from: normal field of views (FOVs) (no collagen); Peri-sinusoidal FOVs (collagen branching around sinusoids); Peri-portal FOVs (collagen branching around portal areas); Portal FOVs (collagen present in portal areas); and Bridging FOVs (collagen present in the form of bridges between main vascular structures portal areas and central vein).

8. A method for determining and scoring liver or hepatic steatosis in a subject, comprising:

providing a liver biopsy slide from said subject;

using a deep learning model:

determining a plurality of field of views (FOVs) on a digitized form of the liver biopsy slide;

selecting a subset of the plurality of FOVs and identifying patterns of hepatic steatosis or liver steatosis within the subset of the plurality of FOVs;

determining a hepatic or liver steatosis (S) score based on the identified patterns of hepatic steatosis or liver steatosis, wherein the hepatic or liver steatosis (S) is scored as follows: S=0 corresponds to less than 5% steatosis, S=1 corresponds to 6%-33% steatosis;

S=2 corresponds to 34%-66% steatosis, and S=3 corresponds to greater than 66% steatosis.

9. A non-transitory computer-readable storage medium comprising instructions which, when executed by a computer, cause the computer to carry out a method comprising:

providing a liver biopsy slide from said subject;

using a deep learning model:

determining a plurality of field of views (FOVs) on a digitized form of the liver biopsy slide;

selecting a subset of the plurality of FOVs, identifying patterns of hepatocyte ballooning and scoring the hepatocyte ballooning patterns;

wherein hepatocyte ballooning (HB) score ranges from 0 to 2 and HB=0 corresponds to no hepatocyte ballooning HB=1 corresponds to moderate hepatocyte ballooning, and HB=2 corresponds to severe hepatocyte ballooning.

10. A non-transitory computer-readable storage medium comprising instructions which, when executed by a computer, cause the computer to carry out a method comprising:

providing a liver biopsy slide from said subject;

using a deep learning model:

determining a plurality of field of views (FOVs) on a digitized form of the liver biopsy slide;

selecting a subset of the plurality of FOVs and identifying patterns of hepatocyte lobular inflammation and scoring hepatocyte lobular inflammation from said liver biopsy slide; wherein lobular inflammation (LI) score ranges from 0 to 3 and LI=0 corresponds to no inflammation; LI=1 corresponds to mild inflammation; LI=2 corresponds to moderate inflammation, and LI=3 corresponds to severe inflammation; and

identifying patterns of hepatocyte ballooning (HB) and scoring hepatocyte ballooning; wherein HB score ranges from 0 to 2 and HB=0 corresponds to no hepatocyte ballooning, HB=1 corresponds to moderate hepatocyte ballooning, and HB=2 corresponds to severe hepatocyte ballooning; and

wherein said HB score and said LI score are added to determine an Activity Index (AI), ranking from 0 to 5, wherein the subject is diagnosed as a NASH subject if AI≥2, with HB≥1 and LI≥1.