Table 2 Associations with the radiomics-based clustering

From: Integration of pre-treatment computational radiomics, deep radiomics, and transcriptomics enhances soft-tissue sarcoma patient prognosis

 

h-RFs

CAE

HSCAE

Ah-RFS

Bh-RFS

Ch-RFs

P-valuea

ACAE

BCAE

P-valuea

A HSCAE

B HSCAE

P-valuea

Size (mm)

   

<0.0001***

  

<0.0001***

  

<0.0001***

Mean ± SD

96 ± 39

131 ± 52

59 ± 33

 

120 ± 51

64 ± 34

 

121 ± 50

64 ± 35

 

Histotype

   

0.0101*

  

0.2498

  

0.2634

M/RC LPS

7/71 (9.9)

13/87 (14.9)

2/67 (3)

 

16/138 (11.6)

6/87 (6.9)

 

15/136 (11)

7/89 (7.9)

 

DD/Pleomorphic LPS

5/71 (7)

16/87 (18.4)

6/67 (9)

 

19/138 (13.8)

8/87 (9.2)

 

20/136 (14.7)

7/89 (7.9)

 

Leiomyosarcoma

4/71 (5.6)

10/87 (11.5)

10/67 (14.9)

 

13/138 (9.4)

11/87 (12.6)

 

11/136 (8.1)

13/89 (14.6)

 

Myxofiborsarcoma

2/71 (2.8)

4/87 (4.6)

7/67 (10.4)

 

4/138 (2.9)

9/87 (10.3)

 

5/136 (3.7)

8/89 (9)

 

Other

19/71 (26.8)

11/87 (12.6)

22/67 (32.8)

 

30/138 (21.7)

22/87 (25.3)

 

30/136 (22.1)

22/89 (24.7)

 

Synovial sarcoma

6/71 (8.5)

6/87 (6.9)

4/67 (6)

 

10/138 (7.2)

6/87 (6.9)

 

9/136 (6.6)

7/89 (7.9)

 

UPS

28/71 (39.4)

27/87 (31)

16/67 (23.9)

 

46/138 (33.3)

25/87 (28.7)

 

46/136 (33.8)

25/89 (28.1)

 

FNCLCC grade

   

0.1559

  

0.0274*

  

0.0168*

Grade I

12/71 (16.9)

7/87 (8)

11/64 (17.2)

 

13/138 (9.4)

17/84 (20.2)

 

12/136 (8.8)

18/86 (20.9)

 

Grade II

29/71 (40.8)

27/87 (31)

20/64 (31.2)

 

44/138 (31.9)

32/84 (38.1)

 

44/136 (32.4)

32/86 (37.2)

 

Grade III

30/71 (42.3)

53/87 (60.9)

33/64 (51.6)

 

81/138 (58.7)

35/84 (41.7)

 

80/136 (58.8)

36/86 (41.9)

 

SARCULATOR groups

   

0.0001***

  

<0.0001***

  

<0.0001***

High Pr-OS

38/71 (53.5)

25/87 (28.7)

42/64 (65.6)

 

47/138 (34.1)

58/84 (69)

 

44/136 (32.4)

61/86 (70.9)

 

Interm. Pr-OS

17/71 (23.9)

25/87 (28.7)

14/64 (21.9)

 

37/138 (26.8)

19/84 (22.6)

 

37/136 (27.2)

19/86 (22.1)

 

Low Pr-OS

16/71 (22.5)

37/87 (42.5)

8/64 (12.5)

 

54/138 (39.1)

7/84 (8.3)

 

55/136 (40.4)

6/86 (7)

 

Semantic radiophenotypes

   

0.0005***

  

<0.0001***

  

<0.0001***

Low risk

33/71 (46.5)

25/87 (28.7)

41/67 (61.2)

 

39/138 (28.3)

60/87 (69)

 

42/136 (30.9)

57/89 (64)

 

High risk

38/71 (53.5)

62/87 (71.3)

26/67 (38.8)

 

99/138 (71.7)

27/87 (31)

 

94/136 (69.1)

32/89 (36)

 

Transcriptomics groups

   

0.1867

  

0.9942

  

0.9129

BRNA

14/38 (36.8)

18/38 (47.4)

9/34 (26.5)

 

24/63 (38.1)

17/47 (36.2)

 

25/65 (38.5)

16/45 (35.6)

 

ARNA

24/38 (63.2)

20/38 (52.6)

25/34 (73.5)

 

39/63 (61.9)

30/47 (63.8)

 

40/65 (61.5)

29/45 (64.4)

 
  1. Data are number of patients with percentage parentheses.
  2. CAE convolutional autoencoder neural network, DD dedifferentiated, FNCLCC French federation of cancer centers, h-RFs handcrafted radiomics features, HSCAE half supervised convolutional autoencoder neural network, LPS liposarcoma, M/RC myxoid/round cells, Pr-OS predicted overall survival, SD standard deviation, UPS undifferentiated pleomorphic sarcoma.
  3. *P < 0.05, ***P < 0.001. Significant results are in bold.
  4. aTests are Chi-square tests for pairs of categorical variables, and, regarding association with size, unpaired Mann–Whitney tests (for CAE and HSCAE) and Friedman test (for h-RFs).
Back to article page